https://bugzilla.wikimedia.org/show_bug.cgi?id=57021

--- Comment #5 from Bryan Davis <[email protected]> ---
(In reply to comment #3)
> (In reply to comment #2)
> > Really? Repeated calls resulted in a large jump in cluster CPU usage...
> 
> Where is the graph? I guess if you call something enough times...

This was almost definitely a case of "calling something enough times". When I
made the patch for 52829 I didn't fully comprehend the size of the WMF
$wgSquidServersNoPurge list and the expected number of times it would be
traversed in the course of a normal request. I have a hunch that if the
configuration had been converted to the 10-15 CIDR ranges that should be able
to cover the production cache clusters at the same time that the change
enabling CIDR usage was rolled out we wouldn't be bothering to look at this.

The follow on patches that have been made for 52829 should help greatly, but
there will still be a performance hit for sites configured with a large number
of $wgSquidServersNoPurge entries. If the large list is primarily/entirely
single IPv4/IPv6 addresses the cost should be constrained to a list traversal
with a single strpos() call per entry.

-- 
You are receiving this mail because:
You are the assignee for the bug.
You are on the CC list for the bug.
_______________________________________________
Wikibugs-l mailing list
[email protected]
https://lists.wikimedia.org/mailman/listinfo/wikibugs-l

Reply via email to