03-03-2010 09:06 AM
03/03/2010 11:14:51 - requesting resource nbucore1.NBU_POLICY.MAXJOBS.sql-sqlmisp
03/03/2010 11:27:27 - granted resource nbucore1.NBU_CLIENT.MAXJOBS.sqlmispSolved! Go to Solution.
03-18-2010 09:09 AM
03-03-2010 11:15 AM
03-03-2010 12:29 PM
03-03-2010 02:59 PM
03-04-2010 06:58 AM
03-04-2010 04:31 PM
03-05-2010 09:18 AM
03-05-2010 01:37 PM
03-12-2010 09:39 AM
03-16-2010 12:07 PM
All-
Thanks to one of the advanced TSEs (Jim Harris, highly recommended!), my case 411-519-814, yielded an EEB for nbrb for 6.5.4.
Mention any of these eTracks and you should be able to get it:
ET1596861, ET1702148, ET1634241, ET1600913, ET1785238, ET1860650, ET1873458, ET1870125, ET1867225, ET1886274, ET1915680, ET1977194
One of these (ET1977194) was actually happening to me where bpduplicates would exit with a 96 error and nbrb would report 2005000 (EMM_ERROR_MediaUnavailable). So that saves me an extra call in and case to Symantec.
I am planning to apply this tomorrow and will report back status.
Brian
03-16-2010 12:27 PM
A couple of other kernel parms that I ended up using that seem to help as well on the master (Solaris 10):
/usr/sbin/ndd -set /dev/tcp tcp_conn_req_max_q 8192
/usr/sbin/ndd -set /dev/tcp tcp_conn_req_max_q0 8192
/usr/sbin/ndd -set /dev/tcp tcp_time_wait_interval 15000
/usr/sbin/ndd -set /dev/tcp tcp_keepalive_interval 3600000
(I put these in /lib/svc/method/net-init to automatically set on boot time, and manually set them while I had NetBackup down)
Definitely disable tcp_fusion (see http://seer.entsupport.symantec.com/docs/306694.htm).
And setup project for kernel parameters (see http://seer.entsupport.symantec.com/docs/307610.htm on how to set them up):
project.max-msg-ids 256
project.max-sem-ids 1024
project.max-shm-ids 256
I didn't bother with shm-max since 8GB (1/4 of system memory) is fine for my master, as it doesn't handle client backups.
Modified /usr/openv/var/global/emm.conf to contain:
NUM_DB_CONNECTIONS=51
NUM_DB_BROWSE_CONNECTIONS=50
NUM_ORB_THREADS=61
And /usr/openv/var/global/server.conf (adding the -gn to be 6 more than NUM_DB_CONNECTIONS), and also up'ed the cache to 75M to start (500M high water mark):
-n NB_<master>
-x tcpip(LocalOnly=YES;ServerPort=13785) -gp 4096 -ct+ -gd DBA -gk DBA -gl DBA -ti 0 -c 75M -ch 500M
-cl 75M -gn 57 -zl -os 1M -o /usr/openv/db//log/server.log
-ud
Hope those can help some of you!
03-17-2010 05:41 AM
03-17-2010 01:12 PM
03-17-2010 06:04 PM
Symantec is on NBRB EEB version 5 after 6.5.4, be sure you get version 6, because they still have a lot of issues with the resource handeling.
03-17-2010 06:26 PM
Not sure if I got version 5 of nbrb EEB (would be nice to know how to tell which version!). but it was for NBU 6.5.4.
cgoliver- I had same problems in 6.5.3.1, they told me there were fixes in 6.5.4... I continued to run into problems (one of which was the one from the ET I mentioned above).
I applied the EEB I just got (no idea which version) and it seems to be running ok for now.
It still takes a heckuva lot more time than I would think when there are only a couple hundred jobs in queue, and looking at the evaluation performance
Looking at evaluation cycle performance, its looking a little better (evaluated 400-500 requests in the 3 minutes we have set for SECONDS_FOR_EVAL_LOOP_RELEASE), whereas it used to be in the 30-40 number of requests.
Also wondered if it has anything to do with how often the catalog backup runs, which cleans out the Sybase transaction log. So make sure you get good hot catalog backups often!
And as for Solaris 9, use the appropriate shared memory, semaphore and message queues kernel parameters (see the NBU 6.5 Perf Tuning Guide, they should be set in /etc/system I believe for Solaris 9), or check with your Solaris admin or google it.
03-18-2010 09:09 AM
03-18-2010 11:09 AM
03-18-2010 01:03 PM
03-18-2010 06:30 PM
03-18-2010 06:30 PM