Monday 14 May 2018

SQL DBA Job Interview Questions with Answers


In order to help job seekers, we are presenting a lot of SQL DBA interview questions with correct answers. If you like our efforts, don't forget to subscribe our You Tube channel where we post several free SQL Server tutorials to help. Its a long post so keep patience.



SQL DBA Interview Questions with Answers
Q. What are the common issues a SQL DBA should deal with as a part of DBA daily job?
Ans:
§  Backup Failure
§  Restore Failure
§  Log Full Issues
§  Blocking Alerts
§  Deadlocks Alerts
§  TEMPDB full issues
§  Disk Full Issues
§  SQL Connectivity Issues
§  Access issues
§  Installation and Upgrade Failures
§  SQL Agent Job failures
§  Performance Issues
§  Resource (Memory/IO/CPU etc.) Utilization Alerts
§  High-Availability and Disaster Recovery related issues

Q. “model” system DB is down and we are trying to create a new database. Is it possible to create a new database when model DB is down?
Ans:
We can’t create a new database when model database is down. SQL Server restart will be unsuccessful when model database is down as TEMPDB creation failed. TEMPDB is created based on model DB configurations, since model DB is down TEMPDB will not be created.
Q. Which operation is faster COMMIT or ROLLBACK? Why?
Ans:
It’s obviously COMMIT is faster than ROLLBACK. Let me explain with an example: Let’s say we opened a transaction and updated 8000 records:
Commit: It’s completed quickly as the operation is already completed and it just marks those dirty pages as committed and when checkpoint happens all those dirty pages will be written to disk.
Rollback: The operation is already updated 8000 records if we need to rollback then again all these updates has to be rolled back which means there are another 8000 log records will be written to LDF which will take time when compared to commit.
Q. What are the different ways available to insert data from a file into SQL Server database table?
Ans:
These are the different ways:
§  BCP
§  BULKINSERT
§  OPENROWSET
§  OPENDATASOURCE
§  OPENQUERY
§  LINKED SERVER
§  IMPORT/EXPORT WIZARD
§  SSIS
Q. What is the scope of different temp objects?
Ans:
Local Temp Table: “CREATE TABLE #TempTable”
Local temporary tables are visible only in the current session, and can be shared between nested stored procedure calls
Table Variable: “DECLARE TABLE @TempTable”
The scope of a local variable is the batch, stored procedure, or statement block in which it is declared. They can be passed as parameters between procedures. They are not subject to transactions and will retain all rows following a rollback.
Derived Table: “SELECT * FROM (SELECT * FROM Customers) AS TempTable”
Is visible to the current query only
Global Temp Table: “CREATE TABLE ##TempTable”
This differs from a #temp table in that it is visible to all processes. When the creating process ends, the table is removed (but will wait until any current activity from other processes is done).
CTE: Common Table Expression
Example CTE:
;WITH YourBigCTE AS
(
big query here
)
SELECT * FROM YourTable1 WHERE ID IN (SELECT ID FROM YourBigCTE)
UNION
SELECT * FROM YourTable2 WHERE ID IN (SELECT ID FROM YourBigCTE)
Scope is next immediate select command. Can be used multiple times within the same CTE command, even recursively, and will last for the duration of the CTE command.
Q. What is the maximum limit of SQL Server instances for a standalone computer? 
Ans:
50 instances on a stand-alone server for all SQL Server editions. SQL Server supports 25 instances on a failover cluster.
Q. What is the cluster node limitation?
Ans:
The number of allowable nodes in the SQL Server cluster depends on your SQL Server version and your Windows Server version. For SQL Server 2008 Standard edition, you can only have two cluster nodes. If you have SQL Server Enterprise, the limit depends on your Windows Server version, 8 cluster nodes for 2003 and 16 nodes for 2008.
Q. Task manager is not showing the correct memory usage by SQL Server. How to identify the exact memory usage from SQL Server?
Ans:
To know the exact memory usage relay on column “physical_memory_in_use_kb” from DMV “sys.dm_os_process_memory”.
Using performance counters also we can find the usage.

Performance object: Process
Counter: Private Bytes
Instance: sqlservr
Performance object: Process
Counter: Working Set
Instance: sqlservr

The Private Bytes counter measures the memory that is currently committed. The Working Set counter measures the physical memory that is currently occupied by the process.
For 64-bit sql servers we can also check the current memory usage using the below performance counter.
Performance object: SQL Server:Memory Manager
Counter: Total Server Memory (KB)

Q. What is the option ”Lock Pages in Memory”?
Ans:
Lock Pages in Memory is a setting that can be set on 64-bit operating systems that essentially tell Windows not to swap out SQL Server memory to disk. By default, this setting is turned off on 64-bit systems, but depends on various conditions this option needs to be turned on.
We must be very careful in dealing with this option. One can enable this after a detailed analysis of current environment.
Following issues may rise when “Lock Pages in Memory” is not turned on:
§  SQL Server performance suddenly decreases.
§  Application that connects to SQL Server may encounter timeouts.
§  The hardware running SQL Server may not respond for a short time periods.

Q. How do you know how much memory has been allocated to sql server using AWE?
Ans:
We can use DBCC MEMORYSTSTUS command to know the memory allocation information. But it’s trick to understand the results.
We can use a DMV called “sys.DM_OS_Memory_Clerks”. Sample query to calculate total AWE memory allocated is “SELECT SUM(awe_allocated_kb) FROM sys.dm_os_memory_clerks”
From 2008 onwards we can get all memory related information using DMV “sys.dm_os_process_memory”.
Q. How to apply service pack on Active / Passive cluster on 2008 and 2012?
Ans:
1. Freeze the service groups on Node A (active node).
2. Confirm all SQL services are stopped on Node B.
3. Upgrade the SQL Server 2008 instance on Node B.
4. Reboot node B.
5. Unfreeze the service group on node A.
6. Fail over the service group to Node B.
7. After the service group comes online, freeze the service group on Node B.
8. Confirm all SQL services are stopped on Node A.
9. Upgrade the SQL Server 2008 instance on Node A.
10. Reboot Node A.
11. Unfreeze the service group on node B.
12. Fail back the service group to Node A.
Q. You find SP is not applied on all the nodes across the cluster. How to apply SP only on required nodes?
Ans:
If you find that the product level is not consistent across all the nodes, you will need to fool the 2005 patch installer into only patching the nodes that need updating. To do so, you will have to perform the following steps:
1.    Fail Instance, Cluster, and MSDTC groups to an unpatched node
2.    Remove any successfully patched nodes from failover candidates of the SQL Server Service of the instance group (do this using Cluster Admin tool)
3.    Run the patch
4.    After the patch installs successfully, add the Nodes removed in Step 2 back to the SQL Server Service of the Instance group
Why do you need to do this? Well when the patch installer determines that not all nodes in the cluster are at the same patch level, a passive node operation will fail and will prevent you from moving forward with any further patching.
Q. How to change the sql server service account in a cluster environment?
Ans:
Method 1: (No failover required)
1. Freeze the service group on active node from cluster administrator and then restart the service.
Method2:
1. Offline the SQL resources
2. Update the service account at SSCM and restart the service as needed
3. Add the SQL resources back to online
Note: Don’t forget to update service account at the remaining nodes on the cluster.
Method 3:
1. Node 2 (inactive node) change the SQL startup account in SQL Studio or SCM
2. Fail over the SQL service group from node 1 to node 2.
3. Node 1 (now the inactive node) change the SQL startup account in SQL Studio or SCM
Q. How to apply service pack on Active / Active cluster Nodes?
Ans:
1. Make a note of all node names (and/or IP addresses), SQL Server virtual names along with preferred nodes. If there are more than three nodes you may need to also take note of possible owners for each SQL resource group. For my example assume that I have a cluster with node1 and node2, SQL1 normally lives on node1 and SQL2 normally lives on node2.
2. To start with a clean slate and ensure any previous updates are completed both nodes should be restarted if possible. Choose the physical node that you you want to patch second and restart that node (in my example node2).
3. Restart the node you want to patch first (node1). This will mean that both active SQL instances are now running on node2. Some restarts will be essential, but you could avoid the first two restarts if you need to keep downtime to a minimum and just fail SQL1 over to node2. The main point here is to always patch a passive node.
4. In cluster administrator remove node1 from the possible owners lists of SQL1 and SQL2. This means that neither SQL instance can fail over to node1 while it is being patched.
5. Run the service pack executable on node1.
6. Restart node1.
7. Add node1 back into the possible owners lists of SQL1 and SQL2 and fail both instances over to node1.
8. Repeat steps 4 – 6 on node2.
9. Add node2 back into the possible owners lists of SQL1 and SQL2 and fail both instances over to node2. Check that the build level is correct and review the SQL Server error logs.
10. Fail SQL1 over to node1. Check build levels and SQL Server error logs
Q. What are the agents in replication?
Ans:
Snapshot Agent: Copy Schema+Data to snapshot folder on distributer. Used in all types of replication.
Log reader Agent: Sends transactions from Publisher to Distributor. Used in transactional replication
Distribution Agent: Applies Snapshots / Transactions to all subscribers’ runs at distributer in PUSH and Runs at Subscriber in PULL. Used in transactional and transactional with updatable subscriptions.
Queue reader Agent: Runs at distributer send back transactions from subscriber to publisher. Used in Transactional With updatable subscriptions.
Merge Agent: Applies initial snapshot to subscribers, from the next time synchronize by resolving the conflicts.
Q. Can we configure log shipping in replicated database?
Ans: Yes
Replication does not continue after a log shipping failover. If a failover occurs, replication agents do not connect to the secondary, so transactions are not replicated to Subscribers. If a failback to the primary occurs, replication resumes. All transactions that log shipping copies from the secondary back to the primary are replicated to Subscribers.
For transactional replication, the behavior of log shipping depends on the sync with backup option. This option can be set on the publication database and distribution database; in log shipping for the Publisher, only the setting on the publication database is relevant.
Setting this option on the publication database ensures that transactions are not delivered to the distribution database until they are backed up at the publication database. The last publication database backup can then be restored at the secondary server without any possibility of the distribution database having transactions that the restored publication database does not have. This option guarantees that if the Publisher fails over to a secondary server, consistency is maintained between the Publisher, Distributor, and Subscribers. Latency and throughput are affected because transactions cannot be delivered to the distribution database until they have been backed up at the Publisher.
Q. What are the best RAID levels to use with SQL Server?
Ans:
Before choosing the RAID (Redundant Array of Independent Disks) we should have a look into usage of SQL Server files.
As a basic thumb rule “Data Files” need random access, “Log files” need sequential access and “TempDB” must be on a fastest drive and must be separated from data and log files.
We have to consider the below factors while choosing the RAID level:
Reliability
Storage Efficiency
Random Read
Random Write
Sequential Read
Sequential Write
Cost.
As an Admin we have to consider all of these parameters in choosing the proper RAID level. Obviously the choice is always between RAID-5 and RAID-10
Q. How to monitor latency in replication?
Ans:
There are three methods.
1.    Replication monitor
2.    Replication commands
3.    Tracer Tokens
1. Replication Monitor: In replication monitor from the list of all subscriptions just double click on the desired subscription. There we find three tabs.
§  Publisher to Distributor History
§  Distributor to Subscriber History
§  Undistributed commands
2. Replication Commands:
Publisher.SP_ReplTran: Checks the pending transactions at p
Distributor.MSReplCommands and MSReplTransactions: Gives the transactions and commands details. Actual T_SQL data is in binary format. From the entry time we can estimate the latency.
Distributor.SP_BrowseReplCmds: It shows the exact_seqno along with the corresponding T-SQL command
sp_replmonitorsubscriptionpendingcmds: It shows the total number of pending commands to be applied at subscriber along with the estimated time.
3. Tracer Tokens:
Available from Replication Monitor or via TSQL statements, Tracer Tokens are special timestamp transactions written to the Publisher’s Transaction Log and picked up by the Log Reader. They are then read by the Distribution Agent and written to the Subscriber. Timestamps for each step are recorded in tracking tables in the Distribution Database and can be displayed in Replication Monitor or via TSQL statements.
When Log Reader picks up Token it records time in MStracer_tokens table in the Distribution database. The Distribution Agent then picks up the Token and records Subscriber(s) write time in the MStracer_history tables also in the Distribution database.
Below is the T-SQL code to use Tracer tokens to troubleshoot the latency issues.
–A SQL Agent JOB to insert a new Tracer Token in the publication database.
USE [AdventureWorks]
Go
EXEC sys.sp_posttracertoken @publication = <PublicationName>
Go
–Token Tracking Tables
USE Distribution
Go
–publisher_commit
SELECT Top 20 * FROM MStracer_tokens Order by tracer_id desc

–subscriber_commit
SELECT Top 20 * FROM MStracer_history Order by parent_tracer_id desc
Q. Can we perform a tail log backup if .mdf file is corrupted?
Ans:
Yes we can perform a tail log as long as the ldf if not corrupted and no bulk logged changes.
A typical tail log backup is having two options, 1. WITH NORECOVERY 2.Continue After Error.
1. WITH NORECOVERY: To make sure no transactions happens after the tal log backup
2. CONTINUE AFTER ERROR: Just to make sure log backup happens even though some meta data pages corrupted.
Q. Let’s say we have a situation. We are restoring a database from a full backup. The restore operation ran for 2 hours and failed with an error 9002 (Insufficient logspace). And the database went to suspect mode. How do you troubleshoot this issue?
Ans:
In that case we can actually add a new log file on other drive and rerun the restore operation using the system stored procedure “sp_add_log_file_recover_suspect_db”. Parameters are the same as while creating a new log file.
Q. Let’s say we have a situation. We are restoring a database from a full backup. The restores operation runs for 2 hours and failed with an error 1105 (Insufficient space on the file group). And the database went to suspect mode. How do you troubleshoot this issue?
Ans:
In that case we can actually add a new data file on another drive and rerun the restore operation using the system stored procedure “sp_add_data_file_recover_suspect_db”. Parameters are the same as while creating a new data file.
Q. Can you describe factors that causes the logfile grow?
Ans:
§  CHECKPOINT has not occurred since last log truncation
§  No log backup happens since last full backup when database is in full recovery
§  An active BACKUP or RESTORE operation is running from long back
§  Long running active transactions
§  Database mirroring is paused or mode is in high performance
§  In replication publisher transactions are not yet delivered to distributer
§  Huge number of database snapshots is being created
Q. How do you troubleshoot a Full transaction log issue?
Ans:
Columns log_reuse_wait and log_reuse_wait_desc of the sys.databases catalog view describes what is the actual problem that causes log full / delay truncation.
§  Backing up the log.
§  Freeing disk space so that the log can automatically grow.
§  Moving the log file to a disk drive with sufficient space.
§  Increasing the size of a log file.
§  Adding a log file on a different disk.
§  Completing or killing a long-running transaction.
Q. Does “Truncate” works in transactional replication?
Ans:
No! As per MSDN blogs information we can’t use TRUNCATE on published database against the published article instead we have to use “DELETE” without where clause.
Q. Consider a situation where publisher database log file has been increasing and there there is just few MB available on disk. As an experienced professional how do you react to this situation? Remember no disk space available and also we can’t create a new log file on other drive
Ans:
Essentially we have to identify the bottleneck which is filling the log file.
As a quick resolution check all possible solutions as below:
§  Resolve if there are any errors in log reader agent / distribution agent
§  Fix if there are any connectivity issues either between publisher – distributor or distributor
§  Fix if there are any issues with I/O at any level
§  Check if there is any huge number of transactions pending from publisher
§  Check if there are any large number of VLF’s (USE DBCC Loginfo)which slows the logreader agent work.
§  Check all database statistics are up-to-date at distributer. Usually we do siwtch off this “Auto Update Stats” by default.
§  To find and resolve these issues we can use “Replication Monitor”, “DBCC Commands”, “SQL Profiler”, “System Tables / SP / Function”.
If incase we can’t resolve just by providing a simple solution we have to shrink the transaction log file. Below are two methods.
To shrink the transaction log file:
1. Backup the log — So transactions in vlf’s are marked as inactive
2. Shrink the logfile using DBCC SHRINKFILE – Inactive VLF’s would be removed
3. If you find no difference in size repeat the above steps 1 and 2
To truncate the transaction log file:
In any case we are not able to provide the solution against the increasing logfile the final solution is disable the replication, truncate the log and reinitialize the subscribers.
1. Disable replication jobs
2. Execute SP_ReplDone procedure. It disable the replication and mark as “Replicate done” for all pending transactions at publisher.
3. Backup the transaction log “WITH TRUNCATE” option.
4. Shrink the log file using “DBCC SHRINKFILE”
5. Flues the article cache using “sp_replflush”.
6. Go to distributor database and truncate the table MSRepl_Commands
7. Connect to replication monitor and reinitialize all subscriptions by generating a new snapshot.
8. Enable all replication related jobs.
Q. Can we add an article to the existing publication without generating a snapshot with all articles?
Ans:
Yes! We can do that. Follow the below steps to publish a new article to the existing publication.
There are two parameters that we need to change to “False”. 1. Immediate Sync and 2. Allow_Ananymous.
Both the fields were set to ON by default. If the Immediate_sync is enabled every time you add a new article it will cause the entire snapshot to be applied and not the one for the particular article alone.
Steps:
1. Change the values to “True” for publication properties “Immediate_Sync” and “Allow_Anonymous” using SP_CHANGEPUBLICATION
2. Add a new article to the publication using SP_AddArticle. While executing this procedure along with the required parameters also specify the parameter “@force_invalidate_snapshot=1”.
3. Add the subscriptions to the publication for the single table/article uisng “SP_ADDSUBSCRIPTION”. While executing this proc specify the parameter “@Reserved = Internal”. Generate a new snapshot which only includes newly added article.
Q. How MAXDOP impacts SQL Server?
Ans:
The Microsoft SQL Server max degree of parallelism (MAXDOP) configuration option controls the number of processors that are used for the execution of a query in a parallel plan. This option determines the computing and threads resources that are used for the query plan operators that perform the work in parallel.
For servers that use more than eight processors, use the following configuration:
MAXDOP=8
For servers that use eight or fewer processors, use the following configuration:
MAXDOP=0 to N
Q. How distributed transactions works in SQL Server?
Ans:
Distributed transactions are the transactions that worked across the databases, instances in the given session. Snapshot isolation level does not support distributed transactions.
We can explicitly start a distributed transaction using “BEGIN DISTRIBUTED TRANSACTION <TranName>”
For example, if BEGIN DISTRIBUTED TRANSACTION is issued on ServerA, the session calls a stored procedure on ServerB and another stored procedure on ServerC. The stored procedure on ServerC executes a distributed query against ServerD, and then all four computers are involved in the distributed transaction. The instance of the Database Engine on ServerA is the originating controlling instance for the transaction.
When a distributed query is executed in a local transaction, the transaction is automatically promoted to a distributed transaction if the target OLE DB data source supports ITransactionLocal. If the target OLE DB data source does not support ITransactionLocal, only read-only operations are allowed in the distributed query.
In order to work with these transactions, make sure below settings are done.
1. MSDTC must be running on all supported instances
2. Choose the option “No authentication required” from MSDTC properties
3. Turn on random options at linked server properties like “RPC”, “RPC Out”, “Data Access” etc.

Q. Can you give some examples for One to One, One to Many and Many to Many relationships?
Ans:
One to One: Citizen – UID
A citizen can have only one UID – A UID can represent only one citizen
One to Many: Customer – Products
A customer can sale number of products – A product can be brought by only one customer
Many to Many: Book – Author
A book can be written by more than one author – An author can write more than one book
Q. What are the phases of sql server database restore process?
Ans:
1. Copy Data: Copies all data,log and index pages from backup file to database mdf, ndf and ldf files
2. REDO: Rollfoward all committed transactions to database and if it finds any uncommitted transactions it goes to the final phase UNDO.
3. UNDO: Rollback any uncommitted transactions and make database available to users.
Q. I wanted to know what are the maximum worker threads setting and active worker thread count on sql server. Can you tell me how to capture this info? What’s the default value for max thread count?
Ans:
We can check the current settings and thread allocation using the below queries.
–Thread setting
select max_workers_count from sys.dm_os_sys_info
–Active threads
select count(*) from sys.dm_os_threads

Default value is 255.
Increasing the number of worker threads may actually decrease the performance because too many threads causes context switching which could take so much of the resources that the OS starts to degrade in overall performance.
Q. Can you explain sql server transaction log architecture?
Ans:
We need to spend some time on this as every SQL DBA must aware of this concept.
Q. See I have an environment, Sunday night full backup, everyday night diff backup and every 45 min a transactional backup. Disaster happened at 2:30 PM on Saturday. You suddenly found that the last Sunday backup has been corrupted. What’s your recovery plan?
Ans:
When you find that the last full backup is corrupted or otherwise unrestorable, making all differentials after that point useless. You then need to go back a further week to the previous full backup (taken 13 days ago), and restore that, plus the differential from 8 days ago, and the subsequent 8 days of transaction logs (assuming none of those ended up corrupted!).
If you’re taking daily full backups, a corrupted full backup only introduce an additional 24 hours of logs to restore.
Alternatively, a log shipped copy of the database could save your bacon (you have a warm standby, and you know the log backups are definitely good).

Q. Full backup size is 300 GB, usually my diff backup size varies between 300 MB and 5 GB, one day unfortunately diff backup size was increased to 250 GB? What might be the reason any idea?
Ans:
Are you the kind of DBA who rebuilds all indexes nightly? Your differential backups can easily be nearly as large as your full backup. That means you’re taking up nearly twice the space just to store the backups, and even worse, you’re talking about twice the time to restore the database.
To avoid these issues with diff backups , ideally schedule the index maintenance to happen right before the full backup.
Q. What is .TUF file? What is the significance of the same? Any implications if the file is deleted?
Ans:
.TUF file is the Transaction Undo File, which is created when performing log shipping to a server in Standby mode.
When the database is in Standby mode the database recovery is done when the log is restored; and this mode also creates a file on destination server with .TUF extension which is the transaction undo file.
This file contains information on all the modifications performed at the time backup is taken.
The file plays a important role in Standby mode… the reason being very obvious while restoring the log backup all uncommited transactions are recorded to the undo file with only commited transactions written to disk which enables the users to read the database. So when we restore next transaction log backup; SQL server will fetch all the uncommited transactions from undo file and check with the new transaction log backup whether commited or not.
If found to be commited the transactions will be written to disk else it will be stored in undo file until it gets commited or rolledback.
If .tuf file is got deleted there is no way to repair logshipping except reconfiguring it from scratch.


SQL Server Database Administrator – Interview Questions
This post “SQL DBA – INTERVIEW QUESTIONS ANSWERS – 2”  helps SQL DBA for interview preparation.
Q. Does TRUNCATE is a DDL or DML and why?
Ans:
TRUNCATE is a DDL command as it directly works with table schema instead of row level. This we can observe by using “sys.dm_tran_locks” while executing the “TRUNCATE” command on a table. It issues a schema lock (sch-M) where as “DELETE” issues exclusive lock. Schema lock issued as it requires resetting the identity value.
Q. What is SQL DUMP? Have you ever dealt with this?
Ans:
When SQL Server is crashed or in hung state due to a Memory/Disk/CPU problems it creates a SQL DUMP file. A DUMP files is a file containing a snapshot of the running process (in this case SQL Server) that includes all of the memory space of that process and the call stack of every thread the process has created. There are two major types of DUMP files:
Full DUMP: It contains entire process space and takes lot of time and space
Mini DUMP: It’s a smaller file contains the memory for the call stack of all threads, the CPU registers and information about which modules are loaded.
Q. We are not able to connect to SQL Server. Can you list out what are all the possible issues and resolutions?
Ans:
This is one of the most common problems every DBA should be able to handle with. Here are the list of possible problems and resolutions. All the problems can be categorized into:
§  Service Down/Issue
§  Network Access/Firewall Issue
§  Authentication and Login issue
§  SQL Server configuration Issue
§  Application Driver or Connection String Issue
Possible Problems:
§  Using a wrong instance name/IP or port
§  Using a wrong user name or password
§  User access might be revoked
§  Trying to access from outside organization VPN
§  SQL Server is down
§  SQL Server is not responding due to high CPU/Memory/Disk I/O
§  Might be a disk full issue
§  Master database might be corrupted
§  User default database may not be online
§  SQL Server port might be blocked
§  We are using named instance name and SQL Browser service is down
§  Using the wrong network protocol
§  Remote connections may not be enabled
§  Network issue with the host windows server
§  Using a wrong client driver (32 bit – 64 bit issues or Old driver using for new version)
§  Version Specific issues, for example an application cannot connect to a contained database when connection pooling is enabled. This issue got fixed in SQL Server 2014 CU1
Resolutions:
The error message itself can tell you how to proceed ahead with the resolution:
§  If possible first thing should be done is, check SQL Server and Windows error log as it can tell us the exact problem and based on that we can determine the possible best resolution.
§  Please cross check connection string information before complaining
§  Cross check hosted windows server and SQL Server are up and running
§  Make sure the SQL login default database is online and accessible
§  Make sure the user access is not revoked
§  Make sure all system databases are up and running
§  Cross check all resource usage that includes Memory, CPU, Disk I/O, Disk Space etc.
§  Try to use IP address and port number instead of instance name, also try with FQDN
§  Try to connect from different possible places/systems to make sure the source system has no issues
§  Check windows server is reachable from remote location using PING
§  Check SQL Server listening on the given port using TELNET <IP> <Port>. Try both from local and remote
§  If the port is blocked add this port to exception list in windows firewall INBOUND rules
§  Make sure SQL Server is configured to allow remote connections
§  If you are also not able to connect then try to connect using DAC and fix the issue by running DBCC commands
§  Try if you can connect using SQLCMD
§  Cross check if there is any recent changes happened in Active Directory security policy
§  Make sure you are using the correct driver to connect to application
§  Cross check if there is any blocking on system process
Q. Can you explain how database snapshots works?
Ans:
Let me explain what happens when we create a database snapshot
§  It creates an empty file known as sparse file for each source database data file
§  Uncommitted transactions are rolled back, thus having a consistent copy of the database
§  All dirty pages will be returned to the disk
§  The user can query the database snapshot
§  Initially the sparse file contains an empty copy of source database data file
§  Snapshot data points to the pages from source database datafile
§  When any modification occurred (INSERT/DELETE/UPDATE) on source database, all modified pages are copied to the sparse file before the actual modification. That means the sparse file contains the old/point in time data (when the time the snapshot taken).
§  Now if you query the snapshot all modified pages are read from sparse file and remaining all unchanged pages are read from the original (source database) data file.
Q. How to know the number of VLF created on a given database log file?
Ans:
Run DBCC LOGINFO; Number of rows returned = Total number of VLF. If it is more than 50 means we need to control the Auto-growth rate. Number of times Auto Grow happens means it increases the number of VLF’s.
Q. Any idea about boot page?
Ans:
In every database there is a page available which stores about the most critical information about that database. This page is called boot page. Boot Page is page 9 in first file on primary file group. We can examine the BOOTPAGE using DBCC PAGE or DBCC DBINFO
Q. Can we hot add CPU to sql server?
Ans:
Yes! Adding CPUs can occur physically by adding new hardware, logically by online hardware partitioning, or virtually through a virtualization layer. Starting with SQL Server 2008, SQL Server supports hot add CPU.
§  Requires hardware that supports hot add CPU.
§  Requires the 64-bit edition of Windows Server 2008 Datacenter or the Windows Server 2008 Enterprise Edition for Itanium-Based Systems operating system.
§  Requires SQL Server Enterprise.
§  SQL Server cannot be configured to use soft NUMA
Once the CPU is added just run RECONFIGURE then sql server recognizes the newly added CPU.
QHow can we check whether the port number is connecting or not on a Server?
Ans:
TELNET <HOSTNAME> PORTNUMBER
TELNET PAXT3DEVSQL24 1433
TELNET PAXT3DEVSQL24 1434
Common Ports:
MSSQL Server: 1433
HTTP TCP 80
HTTPS TCP 443
QWhat is the port numbers used for SQL Server services?
Ans:
§  The default SQL Server port is 1433 but only if it’s a default install. Named instances get a random port number.
§  The browser service runs on port UDP 1434.
§  Reporting services is a web service – so it’s port 80, or 443 if it’s SSL enabled.
§  Analysis service is on 2382 but only if it’s a default install. Named instances get a random port number.
Q: Start SQL Server in different modes?
Ans:
Single User Mode (-m) : sqlcmd –m –d master –S PAXT3DEVSQL11 –c –U sa –P *******
DAC (-A): sqlcmd –A –d master –S PAXT3DEVSQL11 –c –U sa –P *******
Emergency: ALTER DATABASE test_db SET EMERGENCY
Q: How to recover a database that is in suspect stage?
Ans:
ALTER DATABASE test_db SET EMERGENCY
After you execute this statement SQL Server will shutdown the database and restart it without recovering it. This will allow you to view / query database objects, but the database will be in read-only mode. Any attempt to modify data will result in an error similar to the following:
Msg 3908, Level 16, State 1, Line 1 Could not run BEGIN TRANSACTION in database ‘test’ …..etc
ALTER DATABASE test SET SINGLE_USER
GO
DBCC CHECKDB (‘test’, REPAIR_ALLOW_DATA_LOSS) GO
If DBCC CHECKDB statement above succeeds the database is brought back online (but you’ll have to place it in multi-user mode before your users can connect to it). Before you turn the database over to your users you should run other statements to ensure its transactional consistency. If DBCC CHECKDB fails then there is no way to repair the database – you must restore it from a backup.
Q. Can we uninstall/rollback a service packs from SQL Server 2005?
Ans:
No not possible for SQL Server 2005. To rollback a SP you have to uninstall entire product and reinstall it.
For Sql Server 2008 you can uninstall a SP from Add/Remove programs.
Some people are saying that we can do it by backup and replace the resource db. But I am not sure about that.
Q. What is a deadlock and what is a live lock? How will you go about resolving deadlocks?
Ans:
Deadlock is a situation when two processes, each having a lock on one piece of data, attempt to acquire a lock on the other’s piece. Each process would wait indefinitely for the other to release the lock, unless one of the user processes is terminated. SQL Server detects deadlocks and terminates one user’s process.
A livelock is one, where a request for an exclusive lock is repeatedly denied because a series of overlapping shared locks keeps interfering. SQL Server detects the situation after four denials and refuses further shared locks. A livelock also occurs when read transactions monopolize a table or page, forcing a write transaction to wait indefinitely.
Q. SQL Server is not responding. What is action plan?
Ans:
Connect using DAC via CMD or SSMS
Connect via CMD
SQLCMD -A –U myadminlogin –P mypassword -SMyServer –dmaster
Once you connect to the master database run the diagnostic quires to find the problem
Correct the issue and restart the server
Find the errors from sql log using
SQLCMD –A –SmyServer –q”Exec xp_readerrorlog” –o”C:\logout.txt”
A long running query blocking all processes and not allowing new connections
Write a query and put the script file on hard disk Ex: D:\Scripts\BlockingQuery.sql
use master;
select p.spid, t.text
from sysprocesses p
CROSS APPLY sys.dm_exec_sql_text (sql_handle) t
where p.blocked = 0
and p.spid in
( select p1.blocked
from sysprocesses p1
where p1.blocked > 0
and p1.waittime > 50 )
From command prompt run the script on sql server and get the result to a text file
SQLCMD -A – SMyServer -i”C:\SQLScripts\GetBlockers.sql” -o”C:\SQLScripts\blockers.txt”
Recently added some data files to temp db and after that SQL Server is not responding
This can occur when you specify new files in a directory to which the SQL Server service account does not have access.
Start the sql server in minimal configuration mode using the startup parameter “–f”. When we specify –f the sql server creates new tempdb files at default file locations and ignore the current tempdb data files configuration. Take care when using –f as it keep the server in single user mode.
Once the server is started change the tempdb configuration settings and restart the server in full mode by removing the flag -f
A database stays in a SUSPECT or RECOVERY_PENDING State
Try to resolve this using CheckDB and any other DBCC commands if you can.
Last and final option is put the db in emergency mode and run CHECKDB with repair_allow_data_loss
(Note: Try to avoid this unless you don’t have any option as you may lose large amounts of data)
Q. What is your experience with third party applications and why would you use them?
Ans:
I have used some of the 3
rd Party tools:
§  SQL CHECK – Idera – Monitoring server activities and memory levels
§  SQL DOC 2 – RedGate – Documenting the databases
§  SQL Backup 5 – RedGate – Automating the Backup Process
§  SQL Prompt – RedGate – Provides IntelliSense for SQL SERVER 2005/2000,
§  Lite Speed 5.0 – Quest Soft – Backup and Restore
Benefits using Third Party Tools:
§  Faster backups and restores
§  Flexible backup and recovery options
§  Secure backups with encryption
§  Enterprise view of your backup and recovery environment
§  Easily identify optimal backup settings
§  Visibility into the transaction log and transaction log backups
§  Timeline view of backup history and schedules
§  Recover individual database objects
§  Encapsulate a complete database restore into a single file to speed up restore time
§  When we need to improve upon the functionality that SQL Server offers natively
§  Save time, better information or notification
Q. Why sql server is better than other databases?
Ans:
I am not going to say one is better than other, but it depends on the requirements. We have number of products in market. But if I have the chance to choose one of them I will choose SQL SERVER because…..
§  According to the 2005 Survey of Wintercorp, The largest SQL Server DW database is the 19.5 terabytes. It is a database of a European Bank
§  High Security. It is offering high level of security.
§  Speed and Concurrency, SQL Server 2005 system is able to handles 5,000 transactions per second and 100,000 queries a day and can scale up to 8 million new rows of data per day,
§  Finally more technical peoples are available for SQL SERVER when we compare to any other database.
So that we can say SQL SERVER is more than enough for any type of application.
Q. What are the Hotfixes and Patches?
Ans:
Hotfixs are software patches that were applied to live i.e. still running systems. A hotfixis a single, cumulative package that includes one or more files that are used to address a problem in a software product (i.e. a software bug).
In a Microsoft SQL SERVER context, hotfixes are small patches designed to address specific issues, most commonly to freshly-discovered security holes.
Ex: If a select query returning duplicate rows with aggregations the result may be wrong….
Q. Why Shrink file/ Shrink DB/ Auto Shrink is really bad?
Ans:
In the SHRINKFILE command, SQL Server isn’t especially careful about where it puts the pages being moved from the end of the file to open pages towards the beginning of the file.
§  The data becomes fragmented, potentially up to 100% fragmentation, this is a performance killer for your database;
§  The operation is slow – all pointers to / from the page / rows being moved have to be fixed up, and the SHRINKFILE operation is single-threaded, so it can be really slow (the single-threaded nature of SHRINKFILE is not going to change any time soon)
Recommendations:
§  Shrink the file by using Truncate Only: First it removes the inactive part of the log and then perform shrink operation
§  Rebuild / Reorganize the indexes once the shrink is done so the Fragmentation level is decreased
Q. Which key provides the strongest encryption?
Ans:
AES (256 bit)
The longer the key, the better the encryption, so choose longer keys for more encryption. However there is a larger performance penalty for longer keys. DES is a relatively old and weaker algorithm than AES.
AES: Advanced Encryption Standard
DES: Data Encryption Standard
Q. What is the difference between memory and disk storage?
Ans:
Memory and disk storage both refer to internal storage space in a computer. The term “memory” usually means RAM (Random Access Memory). The terms “disk space” and “storage” usually refer to hard drive storage.
QWhat port do you need to open on your server firewall to enable named pipes connections?
Ans:
Port 445. Named pipes communicate across TCP port 445.
Q. What are the different log files and how to access it?
Ans:
§  SQL Server Error Log: The Error Log, the most important log file, is used to troubleshoot system problems. SQL Server retains backups of the previous six logs, naming each archived log file sequentially. The current error log file is named ERRORLOG. To view the error log, which is located in the %Program-Files%\Microsoft SQL Server\MSSQL.1\MSSQL\LOG\ERRORLOG directory, open SSMS, expand a server node, expand Management, and click SQL Server Logs
§  SQL Server Agent Log: SQL Server’s job scheduling subsystem, SQL Server Agent, maintains a set of log files with warning and error messages about the jobs it has run, written to the %ProgramFiles%\Microsoft SQL Server\MSSQL.1\MSSQL\LOG directory. SQL Server will maintain up to nine SQL Server Agent error log files. The current log file is named SQLAGENT.OUT, whereas archived files are numbered sequentially. You can view SQL Server Agent logs by using SQL Server Management Studio (SSMS). Expand a server node, expand Management, click SQL Server Logs, and select the check box for SQL Server Agent.
§  Windows Event Log: An important source of information for troubleshooting SQL Server errors, the Windows Event log contains three useful logs. The application log records events in SQL Server and SQL Server Agent and can be used by SQL Server Integration Services (SSIS) packages. The security log records authentication information, and the system log records service startup and shutdown information. To view the Windows Event log, go to Administrative Tools, Event Viewer.
§  SQL Server Setup Log: You might already be familiar with the SQL Server Setup log, which is located at %ProgramFiles%\Microsoft SQL Server\90\Setup Bootstrap\LOG\Summary.txt. If the summary.txt log file shows a component failure, you can investigate the root cause by looking at the component’s log, which you’ll find in the %Program-Files%\Microsoft SQL Server\90\Setup Bootstrap\LOG\Files directory.
§  SQL Server Profiler Log: SQL Server Profiler, the primary application-tracing tool in SQL Server, captures the system’s current database activity and writes it to a file for later analysis. You can find the Profiler logs in the log .trc file in the %ProgramFiles%\Microsoft SQL Server\MSSQL.1\MSSQL\LOG directory.
§   
Q. Explain XP_READERRORLOG or SP_READERRORLOG
Ans:
Xp_readerrorlog or sp_readerrorlog has 7 parameters.
Xp_readerrorlog <Log_FileNo>,<Log_Type>,<Keyword-1>,<Keyword-2>,<Date1>,<Date2>,<’Asc’/’Desc’>
Log_FileNo: -1: All logs
0: Current log file
1: No1 archived log file etc
Log_Type: 1: SQL Server
2: SQL Agent
KeyWord-1: Search for the keyword
KeyWord-2: Search for combination of Keyword 1 and Keyword 2
Date1 and Date2: Retrieves data between these two dates
‘Asc’/’Desc’: Order the data
Examples:
EXEC Xp_readerrorlog 0 – Current SQL Server log
EXEC Xp_readerrorlog 0, 1 – Current SQL Server log
EXEC Xp_readerrorlog 0, 2 – Current SQL Agent log
EXEC Xp_readerrorlog -1 – Entire log file
EXEC Xp_readerrorlog 0, 1, ’dbcc’ – Current SQL server log with dbcc in the string
EXEC Xp_readerrorlog 1, 1, ’dbcc’, ’error’ – Archived 1 SQL server log with dbcc and error in the string
EXEC xp_readerrorlog -1, 1, ‘dbcc’, ‘error’, ‘2012-02-21’, ‘2012-02-22′,’desc’
Search entire sql server log file for string ‘dbcc’ and ‘Error’ within the given dates and retrieves in descending order.
Note: Also, to increase the number of log files, add a new registry key “NumErrorLogs” (REG_DWORD) under below location.
HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\MSSQL.X\MSSQLServer\
By default, this key is absent. Modify the value to the number of logs that you want to maintain.

Q. Can we track no of transactions  / inserts / updates / deletes a Day (Without using profiler)? If yes how?
Ans:
You could use capture data change or change tracking:
Q. We have 300 SSIS packages those needs to be deployed to production, how can we make it easier / short way to deploy all SSIS packages at once.
Ans:
I would store these as XML based files and not in the MSDB database. With the configuration files, you can point the packages from prod to dev (and vice versa) in just a few seconds. The packages and config files are just stored in a directory of your choice. Resources permitting, create a standalone SSIS server away from the primary SQL Server
Q. We have a table which is 1.2 GB in size, we need to write a SP which should work with a particular point of time data (like snapshot) (We should not use snapshot Isolation as it take other 1.2 TB size)
Ans:
You may want to add insert timestamps and update timestamps for each record. Every time a new record is inserted, stamp it with the datetime, and also stamp it with the date time when updated. Also possibly use partitioning to reduce index rebuilds.
Q. What is RAID levels? Which one we have to choose for SQL Server user databases?
Ans:
Check out the charts in this document. It shows how the disks are setup. It will depend on what the customer wants to spend and level of reliability needed. Raid 5 is common, but see the topic ‘RAID 10 versus RAID 5 in Relational Databases’, in the document below. It’s a good discussion. Raid 10 (pronounced Raid one-zero) is supposed to have the best in terms of performance and reliability, but the cost is higher.
Q. How many datafiles I can put in Tempdb? What is the effect of adding multiple data files.
Ans:
By far, the most effective configuration is to set tempdb on its own separate fast drive away from the user databases. I would set the number of files based on # of cpu’s divided by 2. So, if you have 8 cpu’s, then set 4 tempdb files. Set the tempdb large enough with 10% data growth. I would start at a general size of 10 GB for each size. I also would not create more than 4 files for each mdf/ldf even if there were more than 8 cpu’s. you can always add more later.
Q. Let’s say a user is performing a transaction on a clustered server and failover has occurred. What will happen to the Transaction?
Ans:
If it is active/passive, there is a good chance the transaction died, but active/passive is considered by some the better as it is not as difficult to administer. I believe that is what we have on active. Still, active/active may be best depending on what the requirements are for the system.
Q. How you do which node is active and which is passive. What are the criteria for deciding the active node?
Ans:
Open Cluster Administrator, check the SQL Server group where you can see current owner. So current owner is the active node and other nodes are passive.
Q. What is the common trace flags used with SQL Server?
Ans:
Deadlock Information: 1204, 1205, 1222
Network Database files: 1807
Log Record for Connections: 4013
Skip Startup Stored Procedures: 4022
Disable Locking Hints: 8755
Forces uniform extent allocations instead of mixed page allocations 1118 – (SQL 2005 and 2008) To reduces TempDB contention.
Q. What is a Trace flag? Types of Trace Flags? How to enable/disable it? How to monitor a trace flag?
Ans:
Q. What are the limitations for RAM and CPU for SQL SERVER 2008 R2?
Ans:
Feature
Standard
Enterprise
Datacenter
Max Memory
64 GB
2TB
Max Memory supported by windows version
Max CPU (Licensed per Socket, not core)
4 Sockets
8 Sockets
Max Memory supported by windows version
Q. Do you know about Resource Database?
Ans:
All sys objects are physically stored in resource database and logically available on every database.
Resource database can faster the service packs or upgrades
Q. Really does resource faster the upgrades? Can you justify?
Ans:
Yes, in earlier versions upgrades requires dropping and recreating system objects now an upgrade requires a copy of the resource file.
We are also capable of rollback the process, because it just needs to overwrite the existing with the older version resource copy.
Q. I have my PROD sql server all system db’s are located on E drive and I need my resource db on H drive how can you move it?
Ans:
No only resource db cannot be moved, Resource db location is always depends on Master database location, if u want to move resource db you should also move master db.
Q. Can we take the backup for Resource DB?
Ans:
No way. The only way if you want to get a backup is use windows backup for option resource mdf and ldf files.
Q. Any idea what is the Resource db mdf and ldf file names?
Ans:
§  mssqlsystemresource.mdf and
§  mssqlsystemresource.ldf
Q. Can you elaborate the requirements specifications for SQL Server 2008?
Ans:

Q. What you do if a column of data type int is out of scope?
Ans:
I do alter column to BigInt
Q. Are you sure the data type Bigint never been out of scope?
Ans:
Yes I am sure.
Let’s take few examples and see how many years will it take for BIGINT to reach its upper limit in a table:
(A) Considering only positive numbers, Max limit of BIGINT = 9,223,372,036,854,775,807
(B) Number of Seconds in a year = 31,536,000
Assume there are 50,000 records inserted per second into the table. Then the number of years it would take to reach the BIGINT max limit is:
9,223,372,036,854,775,807 / 31,536,000 / 50,000 = 5,849,424 years
Similarly,
If we inserted 1 lakh records per second into the table then it would take 2,924,712 yrs
If we inserted 1 million (1000000) records per second into the table then it would take 292,471 yrs
If we inserted 10 million (10000000) records per second into the table then it would take 29,247 yrs
If we inserted 100 million records per second into the table then it would take 2,925 yrs
If we inserted 1000 million records per second into the table then it would take 292 yrs
By this we would have understood that it would take extremely lots of years to reach the max limit of BIGINT.

Thank you
Team RedBush