May 19, 2009 at 10:04 am
Good Day all you DBA gurus!
I am having a problem with users getting disconnected from my virtual database server. I also get disconnected at times when I run a dmv query to look at ONE index on a database table.
When disconnected there is generally an increase in packages going out of the server (it can be a difference from 41/sec to 3,500/sec and sometimes up to 9,000 (I had to reboot the server when that happened)).
This all seems to have occurred after our ops team did an in-place upgrade to the server from Windows 2003 Std with 4GB RAM to Windows 2003 Ent with 6GB of RAM. AND we deployed McAfee database to the server and to the entire organization over 2 months time.
McAfee db is on one virtual server, and the application (web service) it communicates with is on another virtual server within the same network and SAN.
I didn't reinstall SQL Server after th upgrade, I just enabled it to use up to 4GB of RAM.
I lost my performance counters after the upgrade, which I now have to manually re-install every time we reboot the server. (not sure if this is symptom of something)
I'm not sure what I can do short of keeping indexes in order and watching out for file growth. This appears to be an IO issue.
I would like some recommendations on any tuning I can do to server configs. I cannot tune queries all of my dbs are 3rd party.
The server was performing within acceptable limits before the upgrade and McAfee - good benchmarks with the exceptional IO taking longer than 15 seconds and some fluctuation in write time but nothing over 50 ms on average. (at times yes, but not on avg) *see below
SERVER SPECS:
Virtual Server on a Blade - (not divided into LUNS) 500GB disk space allocated to it)
RAID 6 SAN underneath - SAS drives (not my choice, but I cannot change this)
1 GB NIC
SQL Server 2005 SP3
Windows Server 2003 Enterprise R2 SP2 *See Note
Three logical drives (C, E, F) C: OS / E: Data / F: Logs
No dedicated disks from the SAN to database blade
One 2GB switch that all blades communicate through to the network
Database Objects: 13
Database Object Size: 20GB (18 of which is McAfee) combined file space used
JOBS RAN DURING THE DAY
Hourly T-Log backups to network share, take less than 5 sec on avg to complete, were ok prior to upgrade and new db.
BENCHMARK STATS:
Pre Stats:
Network Interface Bytes Received / Sent Sec: 6
Network Queue Length: 0
Buffer Cache Hit: 99 %
Disk sec Read/Write/Disk Q Length E (data): .009, .002, N/A
Disk sec Read/Write/Disk Q Length F (log): .003, .001, N/A
Post Stats:
Network Interface Bytes Received / Sent Sec: 44
Network Queue Length: 0
Buffer Cache Hit: 99 %
Disk sec Read/Write/Disk Q Length E (data): .011, .007, 1.086
Disk sec Read/Write/Disk Q Length F (log): .003, .001, .009
Thanks!!!!!!
-S
May 19, 2009 at 10:12 am
What's the question?
May 19, 2009 at 10:15 am
What you want to do actually????
[font="Comic Sans MS"]+++BLADE+++[/font]:cool:
May 19, 2009 at 10:54 am
Hi guys - sorry bout that I got cut off - here it is again:
Good Day all you DBA gurus!
I am having a problem with users getting disconnected from my virtual database server. I also get disconnected at times when I run a dmv query to look at ONE index on a database table.
When disconnected there is generally an increase in packages going out of the server (it can be a difference from 41/sec to 3,500/sec and sometimes up to 9,000 (I had to reboot the server when that happened)).
This all seems to have occurred after our ops team did an in-place upgrade to the server from Windows 2003 Std with 4GB RAM to Windows 2003 Ent with 6GB of RAM. AND we deployed McAfee database to the server and to the entire organization over 2 months time.
McAfee db is on one virtual server, and the application (web service) it communicates with is on another virtual server within the same network and SAN.
I didn't reinstall SQL Server after th upgrade, I just enabled it to use up to 4GB of RAM.
I lost my performance counters after the upgrade, which I now have to manually re-install every time we reboot the server. (not sure if this is symptom of something)
I'm not sure what I can do short of keeping indexes in order and watching out for file growth. This appears to be an IO issue.
I would like some recommendations on any tuning I can do to server configs. I cannot tune queries all of my dbs are 3rd party.
The server was performing within acceptable limits before the upgrade and McAfee - good benchmarks with the exceptional IO taking longer than 15 seconds and some fluctuation in write time but nothing over 50 ms on average. (at times yes, but not on avg) *see below
SERVER SPECS:
Virtual Server on a Blade - (not divided into LUNS) 500GB disk space allocated to it)
RAID 6 SAN underneath - SAS drives (not my choice, but I cannot change this)
1 GB NIC
SQL Server 2005 SP3
Windows Server 2003 Enterprise R2 SP2 *See Note
Three logical drives (C, E, F) C: OS / E: Data / F: Logs
No dedicated disks from the SAN to database blade
One 2GB switch that all blades communicate through to the network
Database Objects: 13
Database Object Size: 20GB (18 of which is McAfee) combined file space used
JOBS RAN DURING THE DAY
Hourly T-Log backups to network share, take less than 5 sec on avg to complete, were ok prior to upgrade and new db.
BENCHMARK STATS:
Pre Stats:
Network Interface Bytes Received / Sent Sec: 6
Network Queue Length: 0
Buffer Cache Hit: 99 %
Disk sec Read/Write/Disk Q Length E (data): .009, .002, N/A
Disk sec Read/Write/Disk Q Length F (log): .003, .001, N/A
Post Stats:
Network Interface Bytes Received / Sent Sec: 44
Network Queue Length: 0
Buffer Cache Hit: 99 %
Disk sec Read/Write/Disk Q Length E (data): .011, .007, 1.086
Disk sec Read/Write/Disk Q Length F (log): .003, .001, .009
Thanks!!!!!!
-S
May 19, 2009 at 2:14 pm
anyone?
May 19, 2009 at 2:18 pm
Disconnections that appear random often have to do with drivers. I'd be sure that your drivers and firmware for the cards are up to date.
May 19, 2009 at 3:05 pm
Thanks, perhaps that is something related to the OS upgrade?
Would this have an impact on a sudden flood of IOs out from the server? That seemed more to me like disk or network io contention?
Viewing 7 posts - 1 through 6 (of 6 total)
You must be logged in to reply to this topic. Login to reply