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.
Yup, seems to be so. The machine is alive (ping ncatlab.org works) so I guess it's the instiki server that's crashed.
Oh, back up now ... well, how's about that for service!
Maybe the hosting service is trying to pre-empt Google Wave and implement its own real-time messaging service!
The Lab seems to be down. I will ask the usual question. Is this just me?
Later: it’s back.
Not just you. I had to restart the web server. It’s an annoying bug that causes this one and one of these days I’ll figure out a workaround. One of these days…
Thanks a .
One of these days…
Bang! Zoom! Straight to the moon!
(Although I doubt that all of you english-types will get that classic reference).
Nope. Lost on me.
I cheated. I Googled.
@Andrew. Just to show them! Did you know that Rupert is 90 today. Complete with his tartan trousers. I note: Much of the landscape in Rupert is inspired by the Snowdonia landscape.
Gosh! I knew he was old but I didn’t know he was that old. We have a Rupert scarf which caused a bit of a ruction this morning as my eldest decided he was too old to wear it so he ended up with my scarf, the youngest then had the Rupert scarf, and I got his scarf: Thomas the Tank engine.
(Did you know that Rupert’s still going strong? Slightly modernised, at least in looks - I’ve not paid enough attention to the Norwegian dubbing to figure out if the modernisation extends to the storylines.)
November 8 1920 according to the Wikipedia page.
The lab seems to be down. (It finds it hard to get up early on a Sunday morning.)
Just out of silly curiosity.. can anyone match wind speeds of 125kph (80 mph)? We had them here Thursday night. At Capel Curig about 12 miles away, a wind speed of 145kph was recorded (91 mph). We survived!
Don’t we all?
It’s back now - well spotted.
Can’t match the wind, but just had three deer in our garden.
There are no deer heer or should I write no dere here! In fact no deer on Anglesey, and in quite a lot of North Wales. We do have wild (or feral) goats and both plants and fish that are more usual in the Alps, plus loads of Ravens. (I’m a bit of twitcher.)
Firefox reports ‘The network link was interrupted while negotiating a connection. Please try again.’. A ‘gentle reload’ did not fix this. But restarting the webserver did.
The Lab seems to be down and this forum is running at a snail’s pace.
The Lab seems to be down
Hopefully one fine day this stops happening. I spent two days preparing nLab notes for today’s seminar. Now chances are I’ll be left without these notes.
I emailed Andrew so keep your fingers crossed.
I emailed Andrew
He’s being bombarded with emails, then. :)
By the way: does anyone of you still make privat copies of the whole Lab from time to time? Andrew told me that there is a reliable automatic backup system operating, but it might still be good to keep more copies.
Does anyone have any news about what is happening?
I just restarted the web server, and it appears to be back.
Thanks, Mike. What happened because it was off for 11 hours?
Beats me. All I know how to do is restart it, not troubleshoot it. (-:
Sorry everyone!
Bad day to be without internet.
Our semester’s about to end. I’ll bump fixing (or at the least circumventing) this bug to top priority.
It is possible to have more people able to restart the web server without compromising things like personal web privacy. If anyone else would be prepared to do this to give a little coverage over the different time zones, I’m sure the steering committee would love to hear from you!
Porblems with nlab again, it seems. Maybe we are under attack ?
Maybe we are under attack ?
I’m not sure whether you’re being serious or not, but just in case you are let me reassure you that it’s nothing like that. Each night certain stuff happens automatically on the nlab server (such as making sure we’re using the most up to date version of instiki). If the nlab happens to be in the middle of something at the exact moment that things change underneath it, sometimes that triggers a web server crash. And it stays crashed until someone says the magic words to restart it (“By the power of Grayskull” or something like that). As the nLab gets used more, the probability of the coincidence happening increases.
So I need to investigate to see if there’s an easy fix, or find a work-around. But both will take a non-trivial amount of time so I’ve been putting them off as - until recently - it’s not happened all that often.
So take it as a sign of our success!
Yes, lots of people use it. Yesterday I was introduced to a topology student; she remarked once that the topic I said something about ’is much discussed on instiki’. She has much more computer experience than usual grad students so she remembers our site by the software :)
Get her name! We could do with a few more technically minded folks around here.
The lab being down, I think that there do need to be more people who can reboot, but then there needs to be some ’protocol’ to ensure that there are no conflicts with several people rebooting in series (or in parallel, sort of concurrent rebooting!:-) )
I would volunteer to be one but am probably too near in time zone to Andrew for it to be that useful. (There is of course, the ’morning person’ versus ’night person’ question. I tend to start up about 6 in the morning but don’t go on that late at night.)
It seems lab is down now.
Why is the lab software so tempermental?
From what Andrew says this will continue to happen… it is perhaps a mark of our ’success’!!! What is frustrating is that none of the REBOOTERs have come online to reboot things.
In an earlier comment someone mentioned pinging. That works so it is the instiki server that is out of action.
Okay, I’m trying a fix for this. Let’s see if this works.
The problem seems to be with the logging system. I’ve configured instiki to use the syslog facility for its log messages. That works fine (and has some definite advantages over its native logging system), except that syslog rotates its logs every day. While the logs are rotating, they can’t be written to. What seems to happen is that if instiki tries to write at the exact moment that the logs are being shifted, it crashes, and brings down the web server with it (but not the whole machine). So what I’m trying now is something called copytruncate which rather than moving the old log out of the way and then setting up a new one, copies the old log and then blanks the new one. There’s a danger of losing a bit of logging that way, but better that than risking the crashes.
We’ll see if this fixes it.
Interesting. Surely the many other sorts of services that use syslog have found a way around that problem? Can instiki not do whatever it is they do?
The Lab was ’up’ when I logged on this morning, but within a few minutes became slow, then fell over and stopped.
Evidently not.
Eric, you are beginning to annoy me on that one. The real problem is that there is just me maintaining this system and whilst I know “more than the average bear”, I’m not a fully qualified system administrator. We would have the same basic problem with mediawiki as with instiki - namely that I only have a finite amount of time and have many things to do.
I don’t know why you would be annoyed when you support the idea in your very same comment. It IS just you supporting Instiki and you have done a tremendous job. No one would ever question that. Least of all me. But that is exactly the point. I don’t go around saying “You Rock” to everyone and I don’t mean it lightly. However, I disagree that maintaining Mediawiki would be anywhere near the order of magnitude effort that maintaining Instiki is. Would Mediawiki suffer the same syslog issue? I doubt it. There is a ton of support for Mediawiki and you would not have to be inventing solutions like this.
I think the “Unthinkable” thread was not a hysterical brainless rant. It seems to me that it addressed some of the issues you are now experiencing and it is worth revisiting. Go read it again. If it annoyed you, sorry, but that is really your problem in interpreting things. I’m on your side.
For all the reasons we’ve already gone over, Mediawiki is not yet a realistic alternative, but I was just bumping the conversation to refresh memories. It may be worth considering alternatives.
Anyway, my focus has turned elsewhere which is why I haven’t been around lately, so that is pretty much all I have to say on the subject. I hope you take it in a positive light.
But I’m not maintaining Instiki! I do nothing to instiki itself. Whenever we want something tweaking in Instiki, I send an email to Jacques, who reads it very carefully, and then tells me what was wrong with what we wanted and how to do it correctly. If we installed mediawiki, I would have to switch from “system administrator” to “active developer”. As was mentioned in the thread you linked to, no-one has particularly worked on integrating MathML with MediaWiki so I’d probably have to write (and maintain) such a plugin. The software underlying the nForum has a very active support base, but that hasn’t stopped me having to write lots of plugins myself, and fixing bugs, to make it work as we want. So whilst MediaWiki might not suffer the same syslog issue, it would suffer plenty of other issues that would need more advanced fixing. Also, it is not instiki that has a problem with log rotation, but Ruby on Rails. And RoR has lots of active support, I’ve no idea of the numbers but I’d expect it to outnumber the development on MediaWiki. The question is merely tracking down solutions and testing them. That takes time, and that’s what I’m referring to in the limited resources above.
The lab is down. No rush for me as I have to type up a lot of other stuff before I can get back to editing the pages that I was fiddling with! Thanks.
I said the following recently in some private email, but may I should say it again in a slightly wider public:
there might be a rationale for trying to find funds for an IT-graduate student who would look after the nLab as part of a project on something like maybe online interactive authoring systems .Somebody who could devote more time to it than Andrew and who would pesonally benefit more from it, say in that somehow it becomes part of his or her thesis.
Myself I don’t see how I would go about organizing such funding or such a student, but maybe it rings a bell with one of you. After all we already have one student whose project is to attach a database to the nLab. Why should there not be a student whose project is the nLab software itself?
Apart from this, I am wondering what’s going on: in the beginning of the nLab the software had been very buggy. We had to reboot once or twice a day on a regular basis. Then it seemed to me that Andrew had pretty much ironed out these problems. Maybe it’s just that last single syslog-bug waiting to be fixed? But why does it strike so frequently just now?
I send an email to Jacques, who reads it very carefully, and then tells me what was wrong with what we wanted
Hm.
Thank you.
Andrew has thankfully fixed something about my public key, so that now I am finally once again able to log into the server… and reboot if necessary.
I’ve just installed something called monit to keep an eye on the webserver. If I’ve configured it correctly, then it should reboot the web server if it spots that it isn’t responding. It checks every minute, so that ought to be fast enough that no-one will notice!
We will keep our fingers crossed. Thanks.
Thanks, Andrew.
Maybe it’s just that last single syslog-bug waiting to be fixed?
AFAIK, it’s just that and the cache bug.
Okay, good. The monit thing does not quite fix the bug the elegant way, but it does provide a good general purpose robustness. So I am happy.
To be clear, monit is something that I’ve been intending to install for quite some time, as it has the potential to keep things on a more even keel generally. So rather than seeing it as a fix for this problem, see it rather as this problem prompting me to put in place some general purpose strategies that are worthwhile anyway.
That’s precisely what I was thinking (and saying, no? :-).
Two other thoughts on bugs or related:
it seems that something happened to “Recently Revised”. It reacts quicker now. It used to take on the order of a minute to appear. Now it appears sort of in a normal time span (at least for me, on my machine).
there are still some big pages on the nLab that take forever to display, at least right after they have been edited. (for instance infinity-Chern-Weil theory introduction and infinity-Lie groupoid)
but I found out the following trick: I go to the files, edit them and then submit. Instead of going for a walk with the dog and come back to the entry afterwards, one can do the following: hit “stop” on your browser. Then submit the edit again . Then it will take instiki just a comparatively very short time to complain “you tried to sumbmit this without editing anything”. Then hit cancel . Then there the entry is, updated as it should be!
This shows that it’s not actually the saving of the entry that takes minutes. But some wrong action that the software tries to take after that.
What happens if, instead of hitting “stop”, you just reload the page (as a show rather than an edit)? In detail:
How long does it take, and is it the newer page?
Just a sec. The lab is down right now. Let’s see if it recovers within a minute.
Okay, there it is again. Seemed to have taken 2 or 3 minutes, but maybe that was some caching on my end.
Okay Andrew, I have checked what you suggested I should check by adding our new arXiv number to infinity-Lie groupoid. Here is the result of the experiment:
I open the page for the first time today. It takes a little more than a minute to appear.
I reload it in another window: this time it appears in a normal time. Same for reloads.
I hit edit and add a shameless plug for our recent preprint. Then I hit “submit”.
Now about two minutes pass in which the window in which I have submitted is waiting. Meanwhile,
I keep reloading the show-view of the page in the second window again and again: during the first minute or so the page in the second window builds up really quickly – but not with the new edit included, but the old version.
Then after that first minute (the first window still sees no sign of anyting happening) also the reload in the second window stalls. Now about another minute passes with nothing at all happening, then both windows display the new updated page.
So in conclusion: it seems to me that going submitstopresubmitcancel causes something to happen that does not happen otherwise. Something good.
Interesting. I think I need to experiment with the log files streaming past in another window to see what’s really going on.
It seems ok for me. Possibly restarted by someone.
By the way, when does anybody know approximately, how many recipients are at the traditional categories mailing list maintained by Rosebrugh ? If it is commercially how costly are mails which are sent to several thousand users ? How much is this of a concern for spammers ? And finally, how many hits Lab has per day ? I am asking with a constructive reason, so please if no having first hand info and guessing, please somewhat informed guesses.
(Wikipedia) Spamming remains economically viable because advertisers have no operating costs beyond the management of their mailing lists, and it is difficult to hold senders accountable for their mass mailings.
But this does not sound right. I know if you have some free email provider than they will charge you if you send enormously many emails. So from some point on, he number of emails may matter.
If you can define “hit” then I can look through the nLab logs and tell you exactly how many there are.
I keep restarting the Lab, but it keeps being slow or unresponsive. Frustrating.
62 To see what is the most useful definition, let me compare the needs with the info from other source first. I will be back to this next week.
Is the Lab down ? I can not get any page after I entered the internet about 5-10 minutes ago.
1 to 65 of 65