Bug 17891 – forum is dog slow

Status
RESOLVED
Resolution
FIXED
Severity
major
Priority
P1
Component
dlang.org
Product
D
Version
D2
Platform
x86
OS
Windows
Creation time
2017-10-10T19:55:54Z
Last change time
2017-10-19T04:13:18Z
Assigned to
No Owner
Creator
Jonathan Marler

Comments

Comment #0 by johnnymarler — 2017-10-10T19:55:54Z
Lately the forms have been loading alot slower, over 10 seconds to load a single page, sometimes it seems to take longer. Today (October 10, 2017) pages are taking about a minute to load each time making the forums almost unusable.
Comment #1 by dlang-bugzilla — 2017-10-10T20:13:13Z
There is some maintenance running on the server right now which is impacting load. Should be resolved soon.
Comment #2 by kinke — 2017-10-10T21:28:35Z
Right now it's particularly bad, but the overall experience has been pretty bad for weeks now. Not just the web forum, but the Wiki as well. CI is also dying because dlang.org can't serve downloads at times. As I stated in http://forum.dlang.org/thread/[email protected], most requests are served in an instant, but there seem to be times where the server takes 10-60 secs to respond, to the point where it's barely usable. If I have multiple pending requests during such a period (a couple of browser tabs), all of them are served at the same time once the server seems to recover. I'd much rather have a constant delay of a few seconds instead of not-that-infrequent-spikes, those are much more annoying. Of the few websites I use regularly, dlang.org leaves the worst impression performance-wise, by far.
Comment #3 by schveiguy — 2017-10-11T17:33:14Z
I have noticed this same behavior on the forums and dlang.org, instantaneous response most of the time, but then all of a sudden a 10 second delay for one request. Is there something to do with the GC running (assuming these are all backed by D code)? That would fit the pattern. If not, then perhaps there is some networking config issue that is causing the problems.
Comment #4 by kinke — 2017-10-13T19:09:51Z
(In reply to Steven Schveighoffer from comment #3) > Is there something to do with the GC running (assuming these are all backed > by D code)? That would fit the pattern. I silently hoped this wouldn't be the case - if that's the reason, it'd be a prime example where the current GC fails badly.
Comment #5 by b2.temp — 2017-10-19T04:13:18Z
It's clearly fixed now.