Happy Friday!
Is it true that a site-bdii will poll resource bdii's & cache data?
If true then queries to site-bdii can be answered very fast fr cache.
OR
Is it true that a site-bdii will poll resource bdii's *only*
when the site-bdii is queried for info?
If true, then queries to site-bdii can timeout if slow resource bdii's.
We've switched from lcgce01 (old SL4 site-bdii) to lcgbdii (SL5 VM bdii)
& it's failing OPS SAM tests with 40-sec timeouts.
The only logical culprit is performance.
There is some load on lcgbdii's VM host (4 cores, 16GB RAM) which is also
NFS server for WN; but I think the real problem is slow resource-bdii
response from 2 CE's that are VM's on another host (8 cores, 16GB RAM).
2 other hardware resources (CE & SE) always show low(=0) response time.
On both VM CE's, figures like this _occasionally_ in /opt/bdii/var/bdii.log:
Time for searches: 54 s
Time for searches: 30 s
Time to update DB: 53 s
Time for searches: 82 s
Time for searches: 48 s
Time for searches: 110 s
Time for searches: 61 s
If the site-bdii *caches* info it should answer instantly from its cache.
But if it doesn't, it'll be at the mercy of any slow resource bdii.
Grateful for Advice, & Pointers to online lucid explain always welcome.
|