You merely adopted the 500. I was born in it, molded by it. I didn't see a fully rendered Discourse page until I was already a man
-
So, could somebody please reset the server again? The 500 erros are getting old really fast.
@Sam , @apapadimoulis , @codinghorror , (@pjh)
Filed Under: I'd take the 2 minutes reboot time over the not really sporadic 500 time, thank you very much
Also: @Sam, if you could share some insight on why this happens right now, that'd be great. (the bots are off) Thanks.
Addendum: things to do during downtime:
- (un)Like things:
- Reply a lot:
- Look at funny graphs:
- get descriptive error messages for editing and then have to retype EVERYTHING!
Addendum2:
@loopback0 said:This is fucking stupid.
- (un)Like things:
-
It's not even consistently broken.
Currently I can post, but notifications are broken. Before I could see posts stream in, but not make a post.No it's just fucked.
Unknown error saving post, try again. Error: 500 Internal Server Error
x3
-
These errors are now almost constant.
It's a fucking joke how badly this site performs; if this was my project, I'd be mortified that it performs this badly.
-
This is fucking stupid.
-
Liked it so much I quoted it in the OP.
Filed Under: Because sometimes a like and a reply are not enough
-
And for a dose of humor:
The unfunny part is how many tries it took to post this. Also, writing in future past tense is weird.
Seriously, I just saw a post stream in but it still won't take my post? And trying to pay this from mobile is fucking awful some it tries giving the editor focus when it pops back up but I have to dismiss the error popup.
It's moments like this that remind me of @Lorne_Kates rant. Also, the fucking auto-complete can show but I still can't post?
And now it does.
-
Also:
and when trying to post it:
-
Can we let them know?
Anyone up on meta.d?
I don't care if I get banned. I'm up for firebombing the shit over there.
Fuck off 500 times Discourse.
I can't even close out the error without clicking due to crappy focus handling.
-
I mentioned Sam and Jeff in this topic and until I know what causes this I am hesitant to post it over at meta.d.
Filed Under: But I'd like any post that adresses this over at meta.d
I can't even close out the error without clicking due to crappy focus handling.
I just hover over the Reply-Button and mash away. Thats the fastest way to generate replies because it takes focus away from the error-message!
Filed Under: The more you know
-
At least it seems to be performing well. It's giving those 500 errors back in hardly any time at all!
-
I just hover over the Reply-Button and mash away. Thats the fastest way to generate replies because it takes focus away from the error-message!
You mean I don't have to click the
OKbutton? Why is it even there then?OOOOH! NEW ONE!
Unknown error saving post, try again. Error: 0 error
-
-
Why is it even there then?
To be fair, I'd be more annoyed if it wasn't there. Clicking on the background is not really intuitive as a way back (if you remember the problems with the enlargened pictures).
Filed Under: One of the few things Discourse does right
-
In my experience, it's pretty consistent that clicking outside a (non-modal) popup dismisses the popup
-
Ok, I'm stupid and angry. I accept both.
Now... has it passed? Maybe?
-
We ARE talking about Discourse here....
Filed Under: Might want to re-evaluate your statement now
-
True, but pressing return should also dismiss it.
-
At some point I think I will write a series of books on how screwed up discourse is.
-
Will you just fork the source-code and print it?
Filed Under: at least distribute the book for free to members of this forum...
-
-
As well as printing meta.d
File Under: Signed copies at that!
-
Discourse is supposed to consider a post "read" if it's on your screen for what, 4 seconds, IIRC. Right now it's taking multiple minutes to mark it read.
-
Right now it's taking multiple minutes to mark it read.
That's just the message bus having gone AWOL because of a 500 storm. :(
-
Oh, I'm well aware of the 500 storm. By the time I was finally able to post that, it seems to be better... maybe? We'll see.
...
Ooh, no 500 error posting this; it went through on the first try. How about this edit? ...
-
Oddly, edits have been almost 100% reliable…
-
edits have been almost 100% reliable…
Not for everyone.
Syntax error popups happened to a few people here during edits.
-
-
Oddly, edits have been almost 100% reliable…
Bullshit.
Syntax errors all over the place trying edits.
edit: by @aliceif
-
Oddly, edits have been almost 100% reliable…
I had one edit not go through whatsoever for ~100 tries so I started another reply in another topic and had the two of them race...
Filed Under: They both lost to my patience
-
Hmm… but then I guess I only did a few edits… not really a representative sample…
-
Oddly, edits have been almost 100% reliable…
Item 4 in the OP:
4. get descriptive error messages for editing and then have to retype EVERYTHING!
-
By the time I was finally able to post that, it seems to be better... maybe? We'll see.
I suspect that there's a problem with the message bus in that when it manages to become disconnected, it really doesn't handle reconnecting all that reliably, and the state of things when it does manage to rejoin is a bit random. I've been seeing issues on and off for months now (IIRC) but usually only when I moved my laptop between networks or otherwise imposed some large-scale disruption on it from outside.
A complete reload will work at reconnecting everything, better than the alternatives, but the state of things when you do that can be strange. After all, who would actively want to see the top of /t/1000??
-
I blame
the immigrants abusing our public health care/t/1000
-
It's baaaaack!
-
Notice how the spikes are so much bigger for t/1000?
-
I rebooted the server.
I wonder if this is from the bots?
-
I wonder if this is from the bots?
You mean all the ones we turned off shortly after midnight UTC, and haven't been running all day, yet the site has been worse than ever, even though weekend traffic is far below weekday traffic?
-
No, I was thinking the ones that weren't disabled....
This seems quite slow though. Suggestions?
EDIT: actually I guess it seems responsive now!
-
Suggestions?
Drop Discourse?In all seriousness, without knowing the details of the setup here, I couldn't even begin to offer suggestions. But it's also clear there's something about the setup that's not right; guess you'll have to discuss this with @PJH and @sam?
-
I'm guessing that there's some kind of awful memory leak still there. @sam has been trying to hunt down those things recently, but he might not have got them all.
(Generic question not aimed at anyone in particular…) Why the fuck has Ruby got these problems in the first place, or has it never been run properly through a leak checker in “I don't want a single byte leaked even from 24 hours of normal operation” mode? It's almost like their devs are so enthusiastic about the cool-of-new that they've forgotten that keeping production services going without intervention is the best of all…
-
I'm guessing that there's some kind of awful memory leak still there
Must be one big-ass bugger of a leak, if it can trigger a 500 storm within an hour or two…
-
Why the fuck has Ruby got these problems in the first place
Uhh.... that would be because Ruby is a pile of shit. But at least it's not JavaScript.
But hey, it's the future. I mean, look, NP-fucking-M got an additional $8M in funding:
How? Why? What the fuckity fuck? I hate modern software development.
-
Uhh.... that would be because Ruby is a pile of shit.
Which then raises the question of why you chose to switch to a forum package written in Ruby… that is, if it was your choice; I wasn't active on CS in the run-up to the switch.
@apapadimoulis said:But at least it's not JavaScript.
Yeah… there's a lot of that in Discourse too
-
Must be one big-ass bugger of a leak, if it can trigger a 500 storm within an hour or two…
It could also be problems with Digital Ocean if they're not isolating the memory requirements of the droplets strongly (because overcommitting is more “cost effective”) and someone else's system is a leaky PoS. That'd be much harder to sort out as it would involve stuff we can't see at all. In fact, let's hope it's not that, given how much of a total bugger it would be.
It's hard to say without seeing the service logs.
-
that would be because Ruby is a pile of shit
Don't I know it. Several of our flagship products are Ruby, and they nearly all need blasted baby-sitting because they're just a bit too inclined to shit in the database in unexpected ways. (The fucked-up string encoding handling is my personal bête noire but some of the other things are even worse.)
-
I rebooted the server.
thanks, let's hope it stays good for now.
I wonder if this is from the bots?
As said by @RaceProUK, most of our bots have been disabled (well, I don't even have one but whatever)
Then again, the bots should not really bring Discourse down. wtdtf is still not the biggest Discourse instance and the others should be handling way more traffic (boing boing, Ubuntu, etc) than the bots are generating.No, I was thinking the ones that weren't disabled....
Most of the active users here are real people right now, afaict.
This seems quite slow though. Suggestions?
You are asking us for suggestions on a highly complex system that uses at least 2 different languages that you yourself almost insult in the next reply on a system we have no access to (and don't even have specs for)... Suprisingly I (personally) have no idea
EDIT: actually I guess it seems responsive now!
It's kinda sad if something that should be the norm is something you have to put inside an EDIT-block....
guess you'll have to discuss this with
@PJH@codinghorror and @sam?PJH only has access to the admin panel and database backups.
but he might not have got them at all.
MTFY
Filed Under: Meh!
-
PJH only has access to the admin panel and database backups.
True, but he's also more active here than any of the Discodevs (with the possible exception of @riking?)… and Jeff will just blame the bots, or t/1000, or servercooties.com, or the phase of the moon, or Kim Jong Il, or something
-
MTFY
He's definitely nailed some of them. Doesn't mean that they're all gone… (Running a production instance under a memory leak checker is pretty much the only way to be sure, and that's not a trivial step to take.)
As I said, careful reading of the logs is required to hunt this. I don't have that access (and don't want it; I want to stay a normal user thanks, as I've too many other things to do already. ;))
-
they're just a bit too inclined to shit in the database in unexpected ways
Which might be a problem here:
INSERT
s seem to suffer more thanUPDATE
s. At least, actions that are connected to them fail more often or rarely, respectively.You are asking us for suggestions on a highly complex system that uses at least 2 different languages that you yourself almost insult in the next reply on a system we have no access to (and don't even have specs for)... Suprisingly I (personally) have no idea
++
-
Uhh.... that would be because Ruby is a pile of shit.
Gee! It's almost as if you made a REALLY FUCKING SHITTY DECISION and are FUCKING INCOMPETENT AT RUNNING A FORUM you piece of shit.
How? Why? What the fuckity fuck? I hate modern software development.
You hate it BUT YOU ALSO ADOPT IT you fucking hypocritical asshole.