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
-
Making sure enough disk space gets provisioned
Note that that being the cause of the symptoms being seen is an assumption on my (and a few others') part.
-
Note that that being the cause of the symptoms being seen is an assumption on my (and a few others') part.
Fair enough, but IIRC it's not the first time it's happened anyway.
-
it has to hold all of jeff's memes
/me hangs his head in shame
But I am only doing my knightly duty!
-
/me hangs his head in shame
But I am only doing my knightly duty!
To be fair, there's probably more Amy Rose images and Picard mega-facepalms.Courtesy of your friendly neighbourhood hedgehog
Which reminds me: we need
:hedgehog:
. And:fox:
.
-
I can't help it anymore. Every time I see this topic I think Was it ever funny?
I don't know. I personally had some fun bashing Discourse for killing itself every other second. If it had only lasted for a short amount of time, we would have had some Discobashing and then gone our merry ways.
Filed Under: Also, don't forget: this is a fun project Alex created. If this doesn't exist anymore due to 500 it means the fun is gone!
Also Filed Under: It might also have been my way of being passive aggressive... while still trying to be as polite as possibleAddendum: Actually, now that I think about it, having negative likes back and negative amounts of posts was pretty cool.
-
By the way, there's something else we may need to start worrying about when it comes to disk space, andI'm increasingly convinced Digital Ocean is not up for the job.
-
Do we do that? I thought we stored images on the server?
-
I'm increasingly convinced Digital Ocean is not up for the job
Maybe. Maybe we just need more hamsters powering it.
-
Looks like we do store them locally, so we won't be worse affected. Even so...
-
Or switch to guinea pigs, they're bigger and I presume have a bit more power output.
-
I'm increasingly convinced Digital Ocean is not up for the job.
where would we move to?
-
I happen to know there is a DailyWTF forum running quite nicely and snappily on AWS.
-
-
I happen to know there is a DailyWTF forum running quite nicely and snappily on AWS.
as do i.
but do you think that Discourse would work better on AWS than DigitalOcean?
I'd be happy to chip in a little cash to move webhost if it would help, but i'm don't know any webhosts other than DO for VPS hosting and DO is more than sufficient for the servers i do run.
of course none of them run discourse...
-
I happen to know there is a DailyWTF forum running quite nicely and snappily on AWS.
I was going to make a snarky comment about that forum running snappily only because it was read-only. Then I thought about it and realized that Discourse probably runs like shit whether it's read-only or not. So, um, I guess you win.
-
but do you think that Discourse would work better on AWS than DigitalOcean?
I have no clue.
I do know that on AWS you can control your VM's RAM/CPU speed and disk space/disk speed separately, so if the problem is in one area and not the other, you don't have to pay to upgrade both.
I know virtually nothing about DigitalOcean, though.
I should also say in fairness, looking at DigitalOcean's price list, the AWS server might end up costing a couple more bucks a month for equivalent capacity.
-
I was going to make a snarky comment about that forum running snappily only because it was read-only.
It's read-only because it has read-only access to the database. It's still chugging all those features (including the tag cloud, which is just hidden with display: none.) Read-only != static HTML.
So, um, I guess you win.
I always win.
-
-
I do know that on AWS you can control your VM's RAM/CPU speed and disk space/disk speed separately,
that is a feature i wish DO had. but then given the price of disk storage and how in lockstep ram/cpu tend to be for resource contention (usually if you have contention in one of them increasing both will give more relief than increasing just one..... usually) that's not that bad of a thing, and it does massively simplify things for DO allowing them to better plan provisioning.
DO also has droplets with dedicated RAM/CPU by default. you are assigned your ram outright and no one else can use it so there's no overcommit like you can get on AWS (not that AWS does that often)
I should also say in fairness, looking at DigitalOcean's price list, the AWS server might end up costing a couple more bucks a month for equivalent capacity.
that's what made me pick DO in the first place. it was cheaper and i do my own sysadmin work. no support for windows hosts though so should i need one of those i'll need a different host.
If i knew more about the droplet we're running on i might be able to get some improvement in perf, but i doubt it. we're probably just underprovisioned.
good news is we should be able to reprovision in place now, including disk. DO added that recently.
-
no support for windows hosts though so should i need one of those i'll need a different host.
AWS ;)
-
-
Looking at some of the posts on meta, scaling Discourse out (multiple front-end instances) on Amazon is now pretty difficult since you can't just build a master AMI, point it at your S3 and RDS accounts and *boom* there it is. You have to have some sort of complicated dance with glusterfs and unionfs and stuff since it expects all resources to be "local". Scaling up (one bigger instance) is still possible, but that's difficult to do automatically and isn't resilient. Not that we care, since we don't use any of those features anyway, but if you're used to AWS and expecting those features to work properly... well, with Discourse, they won't.
Digital Ocean is silly, in my opinion; it's good for things like Bees with Machine Guns, where you need a lot of on-demand power, but it doesn't look suited for stable commitments. Probably oversold and underprovisioned, based on reports from other customers.
vNucleus will end up costing us a little extra if we go with their unmanaged plans ($41 for 4GB RAM / 200GB SSD disk / "6 core", versus $40 for 4GB RAM / 60GB SSD disk / "2 core"), and we'll get some help setting Discourse up, but then we'll be back to our usual "courtesy" support from Sam and scorn from Jeff. Alternatively, we cold go with their managed plan and pay a pretty hefty amount, but then we'll be getting actual support with actual SLAs and such.
NearlyFreeSpeech indicated that it would probably be impossible on their platform, and in any case wouldn't be cost-effective.
Azure is a possibility, but to provision essentially what DO claims to be dedicating to us now would be a 50% price bump, soo...
-
At this point I think hosting from an old Core 2 and cable modem in one of our basements would be an improvement in stability.
-
Maybe there are some good hosting providers in Wisconsin ...
-
vNucleus will end up costing us a little extra if we go with their unmanaged plans ($41 for 4GB RAM / 200GB SSD disk / "6 core", versus $40 for 4GB RAM / 60GB SSD disk / "2 core"), and we'll get some help setting Discourse up, but then we'll be back to our usual "courtesy" support from Sam and scorn from Jeff. Alternatively, we cold go with their managed plan and pay a pretty hefty amount, but then we'll be getting actual support with actual SLAs and such.
if it would give us better result i'd happily pay for an instance of their Self-Managed 800 series or kick in a benjamin a month for one of their managed plans.
I'll even provide the maintenance myself. gratis. just don't expoect a SLA as i do have a day job and y'all are great at breaking the site when i'm on me lunch break
-
At this point I think hosting from an old Core 2 and cable modem in one of our basements would be an improvement in stability.
I actually have a nice server that isn't being used, but I refuse to run Discourse on it. I do not want to bring shame to my family.
-
We'd actually be okay on a 400; we're only using 2GB of RAM (plus 1.8GB cached from disk). The current issues seem to be more related to PostgreSQL being silly ... somehow...
-
IIRC, part of that is because all the indexes are optimised for MySQL
-
We'd actually be okay on a 400; we're only using 2GB of RAM (plus 1.8GB cached from disk). The current issues seem to be more related to PostgreSQL being silly ... somehow...
well then. we have offer of server payment that will all but cover the managed 400 level with the remainder being what alex was paying for the previous, smaller droplet...
@apapadimoulis, @sam what do you say, do we want to give it a try moving provider or not?
I'm happy to help and put cash on the line to help, but i don't want to spend it without good reason.
-
I'm happy to help and put cash on the line to help, but i don't want to spend it without good reason.
I certainly don't mind chipping in a bob or two a month also
-
AWS
Azure is better for Windows servers. Possibly for Linux servers, too. (Then again, it's had a couple high-profile downtime incidents recently...)
Evaluate all your options. I use AWS because that's what I'm used to, not because it's necessarily better.
-
Whatever ends up happening, it's now a solid 5 seconds between me clicking the speech button icon and actually seeing my notifications.
(Which is stupid, BTW. Why is the back-end telling the front-end "hey this user has 4 notifications", without actually telling it what the notifications are! Like, you know the user's going to click the damned speech bubble, pre-load that fucker!)
-
-
The current issues seem to be more related to
PostgreSQL being silly ... somehow...poorly written queries and potentially missing indexes.FTFY
-
What @abarker said. I somehow doubt we're managing to bring down Postgres with amount of traffic and content we produce.
-
Well yes, but if we abide by Blakey's prohibition on "adhering to the King's enemies in his Realm, giving them aid and comfort" making the queries themselves better is a capital offense. Ensuring the database engine itself is set up properly, and the database has appropriate indices for these poorly written queries, at least mitigates some of the stupid.
-
I don't have much experience with Postgres, so at first when we started having issues, I thought "wow how can so many people be using postgres when it is this terrible." Then I did some reading on it and realized it wasn't postgres that has issues--it was Discourse's usage of it.
-
Postgres is used by all sorts of people; just ten seconds, and I find out BASF uses it. And they're a holy-fucking-shit big multinational.
-
Yeah, my $employer uses Postgres for some of the non-critical stuff around here (we're mostly an Oracle shop, though -- our management wants someone to hold accountable if a database hotfix decides to halt trains for a day instead of, you know, fixing things).
-
Well yes, but if we abide by Blakey's prohibition on "adhering to the King's enemies in his Realm, giving them aid and comfort" making the queries themselves better is a capital offense.
I don't care. He can huff and puff all he wants. The forum needs to be stabilized. That means properly identifying the root problem and fixing it.
And based on posts of his I've seen, his problem isn't so much the software. He has … other concerns.
-
What are you talking about, and why is my name on it?
-
Well... if we get an instance where Discodevs have no access... we can sell the fix?
Ok, that would be us being dicks. But I can at least pretend.
-
That means properly identifying the root problem and fixing it.
Root problem Identified: Discourse.
-
BASF uses it. And they're a holy-fucking-shit big multinational.
That is not a barrier to 'ing.
-
Giving the DiscoDevs QA in the form of fixes for their broken queries would be treason, as you've repeatedly made blatantly clear?
-
I happen to know there is a DailyWTF forum running quite nicely and snappily on AWS.
But isn't it read-only?
-
-
:slowpoke.tif:
-
-
Giving the DiscoDevs QA in the form of fixes for their broken queries would be treason,
Oh. I won't do it. Other people can do what they like. I'm not a dictator.
Calling it "treason" is ridiculous.