Happy Friday!
> Steve Jones kindly documented this here:
>
> https://www.gridpp.ac.uk/wiki/Example_Build_of_an_ARC/Condor_Cluster#Patch_for_BDII_Job_Count_Breakdown
Very good page, thank you!
Speaking of ARC-CE & how to do things: our ARC-CE didn't have its DN in
GOC-DB until late May 2016, so publishing failed & failed. It also didn't
have archiving turned on but it was THOUGHT to be on since all the job
records & etc data were still on it! (But that was because publishing was
failing & failing daily.)
Once publishing worked, all the job data was deleted after every successful
publishing. THEN we realized archiving was not configured on it.
6 July, archiving was configured on ARC-CE. But it looks like stuff was
still getting deleted after a while (unknown how - sorry, I'm 99.9%
unfamiliar with ARC-CE anything). In backup indices, files timestamped 7
July are seen, but have been deleted by don't know what.
However, publishing was working (APEL getting data from it.)
On 19 July ARC-CE was brutally shutdown due to machine room cooling failure.
It was rebooted 21 July. From that date files accumulate in /var/run/arc/urs;
the oldest is timestamped 21 July (dunno what happened to previous!),
accounting publishing is still working ok, and NOTHING appears to be being
deleted from /var/run/arc/urs; late July there were >13,000 files in
/var/run/arc/urs, then 20,000 & now >42,000 files in /var/run/arc/urs
What is supposed to trim files from /var/run/arc/urs & where could it be
configured as to how long to keep files, or how many to keep in
/var/run/arc/urs?
I have glanced thru Aug 2016 edition of arc-ce-sysadm-guide.pdf & can't see
any such info, ditto glance thru http://wiki.nordugrid.org/wiki/Accounting
Very grateful for advice+pointers!
|