Want to take part in these discussions? Sign in if you have an account, or apply for one below
Vanilla 1.1.10 is a product of Lussumo. More Information: Documentation, Community Support.
I'm not trying to edit or anything crucial, but the nLab seems to have gone down completely.
Its back!
Thanks, yes. I have restarted it.
Bleugh. This seems to be the "Help, something changed on the disk and I can't cope so I'm shutting down completely" issue which I have no idea why it happens as it really ought not to do so. This is now the third (by my count) occurrence.
The n-Lab seems to be very sick this afternoon. I think it has just crashed. It was similarly sick this morning (7.30 o'clock UK time) but then that seemed to clear up.
It seems to go down at the weekend. Is this due to maintenance ?
For the record, it's up at the moment.
It seemed to come up late yesterday (9 hours ago).
My instinct would be to say that it is not that it crashes more at the weekend, it's more that there's fewer lab elves around to restart it.
That said, it does seem to happen more often at weekends than weekdays. There appear to be two kinds of "crash": either some idiot decides that they are going to download the entire nlab in one go, tying up the server and grinding it to a halt. Or there's some mystery bug whereby instiki (or one of the components running it) doesn't like it when the logs rotate, or instiki gets updated, while it's trying to do something. For some reason, this has happened mainly on weekends.
I'm not sure what yesterday's is down to, I'll investigate.
It really would help if we had someone who knew what they were doing working for the nlab! Yet again, I've fallen into the trap of thinking that something that is set up correctly on Jacques' installation of Instiki will automatically be set up correctly on ours. In this case, it was the 'robots.txt' file. There is one in the base directory of the instiki installation so I assumed it was set up correctly. Now that I've read the official docs, I see that it's not a trivial job to make this automatic, and it's not even knowable (by instiki) where this file ought to be, so it's up to the Responsible System Administrator to ensure that it is in the right place.
I've now done that. So, for each web on the nlab, robots are respectfully requested not to request addresses that begin with:
feeds
export
revision
show/diff
print
tex
pdf
list
If anyone with a personal web would like robots to traverse those links, please let me know (and explain why). If anyone has any other possibilities of things that I may have overlooked, also please let me know.
Incidentally, there best I can come up with for the slow-down yesterday was a spike in traffic. Slurping through the logs and looking at the number of requests per minute, then we have considerable spikes at 07:34/5 and at 07:42/3. Then a massive spike at 09:37.
There wasn't an actual crash over the weekend, though. The logs show no record of it needing restarting. I think it was just a vast quantity of requests. The 'robots.txt' file should tame that a little.
Thanks, Andrew!!
I echo that. The forum is extremely useful and could not work without the contribution of Andrew. Thanks.
Thanks! I appreciate being appreciated!
I also echo that. Thank you Andrew! I am ever more amazed at what you guys have created here.
The nLab server seems to be off! I got through briefly a few minutes ago, but then got
Error 502 Ray ID: 453dd8e56f83bc50 • 2018-09-02 06:16:21 UTC Bad gateway
I’m on it. This is caused by something I am trying to fix. Unfortunately the Cloudfare layer in front of the nLab is very unhelpful in this regard; I do not have control over it.
Just to confirm that the nLab will be unstable until further notice, so best to avoid editing until I make an announcement here.
Should hopefully be stable for editing now. Need to look further into the asynchronous stuff, but this is not critical. Will return to it later, cannot look into it further just now.
I will be doing some further work/testing on the asynchronous stuff now. This may cause disruption; if editing, please make sure to save your edits before you submit!
Should be stable again now. Asynchronous stuff still not working properly in general, but it is not a critical problem (in many cases it should not be noticeable). Will return to it when I get the chance.
Just to say that right now all requests to the nLab produce Error 502.
Sorry, false alarm, all good now.
Have switched off the asynchronous rendering completely until I have time to work on it (hopefully later today). Have some ideas of how to get around the problems, but no time just now.
Thank you for mentioning this, and apologies to all for the inconvenience. Yes, Urs alerted me and I was able to bring the server back to life. I was at work and not available to debug the root cause unfortunately. There have been a couple of other occasions recently where we’ve had brief issues of a similar kind. The planned fix for this kind of thing is the move to the cloud, where a completely different infrastructure will be involved,
Is the nLab down?
Is the nLab down? …. It is back up it seems.
The nLab has been going down each day at around noon CET for the last ~2 weeks. It always came back within ~5-30 minutes (always right before I finished writing an email to Richard :-).
So it looks like when it goes down again tomorrow and the following days, there is good reason to expect that it will come back by itself in a few minutes. Still, reason to look forward to having the nLab migrated.
1 to 28 of 28