Hello!
We had a nasty incident today -- a production application slowed down to a
crawl. There was nothing wrong on the application side -- low CPU-load,
low
RAM-consumption, but the DBA was at a loss as to what could possibly be
wrong with the Sybase server -- there was no lock-contention, the number
of
active sessions was low, etc.
Finally the application-vendor asked for the output of "sp_monitorconfig
'all'" (a procedure, which we -- the application people -- aren't even
allowed to run).
In the output there was a line:
Name Num_free Num_active Pct_act Max_Used Num_Reuse
number of open indexes 0 60000 100.00 60000 9916
As soon as the limit was bumped up the application regained performance.
The first question, of course, is:
Why was not reaching the limit logged somewhere for the DBAs to see?
But more im****tantly, 60K open indexes is INSANELY large for our app. We
never have more than 100 simultanious Sybase connections and the total
number of tables in the only database on the server (other than master,
and
tempdb) is under a thousand.
This suggests either an outright leak of the open indexes, or simply a
failure to decrease the counter under some circumastances... Is this a
known problem? Is there fix for the 12.5.4.x branch?
According to the vendor, they have seen this before but the only known (to
them) remedy was restarting Sybase...
Thanks!
-mi
--
http://libpipe.com
- ease your Sybase backups
|
