Log in

Forum:Intermittent and inconsistent performance and apparent downtime

From PathfinderWiki
Forums: Workshop > Intermittent and inconsistent performance and apparent downtime

I've received several complaints from other admins about the wiki's performance, particularly for editing. In most cases, I've not been able to reproduce these performance issues, and not all admins have been able to reproduce them consistently.

To attempt to diagnose these issues, I've restarted individual services that the wiki uses, as well as the entire server, which led to a few minutes of intermittent downtime between June 2 and today.

If you have problems accessing this site, please let us know by commenting here, at User talk:Oznogon, or by emailing me at garrett+pfw at oznogon dot com.

We were down for about 2-3 hours today - the wiki's facebook page got messages from 3 users mentioning this. One said he has not been able to get on in over 2 weeks. even has the 'down' hit from one of the users this morning

I'm not on Facebook, and it was up when I went to bed last night and as soon as I got up this morning, so I'd really, really appreciate details about the timing and locations of the reported downtime.

The messages all came in between 10am and 11am EST. I was having intermittent issues between 9 and 10 myself, but it seems everyone was getting server not responding errors during that 2 hour window between 10am and noon EST (the messages were from Chile, Michigan, and Massachusetts, and I'm in Connecticut). Unfortunately Edge doesn't give detailed error codes, and I didn't try on chrome or firefox at the time: "Hmm, we can't reach this page. Make sure you’ve got the right web address:" Is all you could get out of it.

...and was down again this morning UK time, so some 12+ hours ago from date stamp of this post. There is an uncanny coincidence building up. Does anyone know why is so poor of late? Have not been on wiki for a bit so cannot comment on our wiki performance recently.

Just looked at and there are very few comments and nothing official about the outages of late, which I find odd.

It's purely coincidence from the standpoint of what we can control. We share no infrastructure with Paizo, and whatever issues they're encountering are not aligned with ours ( is down right now, but we're not and haven't been all day).

In fact, there are no indications on our server that it was down, under significant load, or unable to serve users, and there were no reported outages of our DNS services, so I continue to suspect that our problems are external to the wiki itself and lie with the provider.

Update: As of June 16, the PathfinderWiki server restarts daily at 09:00 UTC (2 a.m. Pacific, 5 a.m. Eastern), near our lowest point in global site traffic. This restart generally takes less than 1 minute. These restarts are a workaround to a server limitation that leads to performance degradation over time, which increases server response times and can lead to the intermittent availability issues we experienced in May and June. We're trading a small amount of downtime for improved response times, and they've paid off in analytics so far, with our average site response time down from over 750ms to 275ms-300ms. (For context,'s average response times over the last 24 hours are around 345ms.)

While logged in, you should see a banner starting at around 08:00 UTC as a warning of an impending restart. You can continue to work normally, but be aware that attempting to save an edit during the 30-60 seconds of a server reboot might fail.

Argh, this is such an old-fashioned forum.

Anyway, good news: I have been hammering away quite a bit and have had zero performance problems. Normally, over recent months, things would have got quite bad in the time I have been editing. Indeed, I'd say the performance was actually noticeably faster than on a good day before the upgrade. Many thanks Oz.