Red Hat Bugzilla – Bug 853270
Last modified: 2014-12-07 20:13:31 EST
I suggest that log-delete handle authentication within the code itself rather
than rely perhaps on running kinit from the shell.
Also, log-delete is a single threaded process. Making it multi 'threaded' would
make it run much quicker.
*** Bug 853836 has been marked as a duplicate of this bug. ***
(In reply to comment #1)
(In reply to comment #2)
> *** Bug 853836 has been marked as a duplicate of this bug. ***
Setting back to ASSIGNED since the Gerrit review in comment 1 is for a different bug.
(In reply to comment #0)
> Also, log-delete is a single threaded process. Making it multi 'threaded'
> make it run much quicker.
Maybe we should open a separate bug about this. I would like to see some hard data about where the bottlenecks actually are before we do anything complicated like parallelise the DELETE requests.
It seems it might be easier to parralise the dav requests in python-requests as it seems to be thread safe (unlike urllib2).
I just ran a test but realised that it's not going to be very reflective of the real world. On average dav deletes were taking about 0.03 seconds. The server and the dav server are on the same VM though, So I don't imagine you wouldn't get much faster.
However it still took ~60secs to delete all the logs for a single job, whilst it took 4 seconds to delete the job from the database.
Tomorrow I'll try testing more jobs with fewer tasks.
I of course meant to say they are VMs on the same host
Seems like this isn't needed.