AWS issues
-
@dangeRuss I have doubts that this software is ready for multihoming.
-
@Weng said in AWS issues:
What the fuck is a CPU credit and why has nobody ever mentioned this in my umpteen discussions about using AWS at work?
I was under the impression that if my AWS VM was on, it's resources were fully allocated for my use.
Unless there's some janky cutrate tier I scrolled right past.
@Weng said in AWS issues:
What the fuck is a CPU credit and why has nobody ever mentioned this in my umpteen discussions about using AWS at work?
I was under the impression that if my AWS VM was on, it's resources were fully allocated for my use.
Unless there's some janky cutrate tier I scrolled right past.
The T2 tier gives you a certain % of the CPU cores as a baseline then issues you X number of CPU credits per hour. Those credits are burned in response to need for higher capacity, giving you more performance out of the machine, but only when necessary.
-
@e4tmyl33t said in AWS issues:
The T2 tier gives you a certain % of the CPU cores as a baseline
If we're going to stay scalable, our baseline specs should be able to handle our baseline amount of traffic. Just saying.
-
@Yamikuronue said in AWS issues:
@e4tmyl33t said in AWS issues:
The T2 tier gives you a certain % of the CPU cores as a baseline
If we're going to stay scalable, our baseline specs should be able to handle our baseline amount of traffic. Just saying.
As the layperson in this, I would imagine that that's how it's supposed to work. But there I go making sense again...
-
@e4tmyl33t said in AWS issues:
making sense again...
-
@Yamikuronue A reserved m4.large is a measly 600 bucks a year. An m3.medium (which is basically a full time t2.large) is half that.
Looking at these adorable baby t2's, there is no fucking way I'd ever use them for production.
-
... I want to know how the fuck a t2.micro with Windows is cheaper than a t2.micro with RHEL or SLES.
-
@Weng said in AWS issues:
... I want to know how the fuck a t2.micro with Windows is cheaper than a t2.micro with RHEL or SLES.
Hmm
windows
t2.micro 1 Variable 1 EBS Only $0.018 per HourRHEL
t2.micro 1 Variable 1 EBS Only $0.073 per HourSLES
t2.micro 1 Variable 1 EBS Only $0.023 per HourWell RHEL costs money too, and not sure about SLES.
On the other hand RHEL is probably licensed per instance, but with windows you can probably run a shitton VMs with a single datacenter license (on a single physical box).
-
AWS pricing is complicated, why did it move from ocean?
-
@dangeRuss At AWS the t2 or m4 part indicates the physical box. The biggest instance for a particular class is just a hair short of "the entire box".
Since the same pricing pattern continues all the way through the t2 series... Wtf?
Unless the t2 is a special case, because even the biggest instances are fucking teeny. I'll have to see if my Amazon insider buddy knows anything about t-series hardware.
-
@Weng
t2.large 2 Variable 8 EBS Only $0.134 per Hour
That's the biggest they have, a whopping 2 CPU.Since they have
m4.16xlarge 64 188 256 EBS Only $7.862 per Hour
I assume they run the micro instances on giant boxes as well. So even if they are just running 32 instances of t2.large on a physical box, that's still 32 licenses they get out of a single windows datacenter license.
-
It seems like this needs sorting, but in the immediate term, would it help if us regular users all promised to stay off the forum for a day to let the CPU credits recover?
-
@CarrieVS said in AWS issues:
It seems like this needs sorting, but in the immediate term, would it help if us regular users all promised to stay off the forum for a day to let the CPU credits recover?
# echo 127.0.0.1 what.thedailywtf.com >> /etc/hosts
-
@CarrieVS said in AWS issues:
It seems like this needs sorting, but in the immediate term, would it help if us regular users all promised to stay off the forum for a day to let the CPU credits recover?
I'd much rather have the regulars using the forum than the search engine bots.
-
@Weng said in AWS issues:
Unless the t2 is a special case, because even the biggest instances are fucking teeny. I'll have to see if my Amazon insider buddy knows anything about t-series hardware.
@dangeRuss said in AWS issues:
@Weng
t2.large 2 Variable 8 EBS Only $0.134 per Hour
That's the biggest they have, a whopping 2 CPU.Basically, t-series boxes are underprovisioned. You only own a slice of those one or two cores, but as you go up in size, your slice of the CPU time increases. So t2.nano has 3 CPU credits and access to 1 core, but those 3 credits mean you can only max out the core for 3 minutes an hour before you start digging into your credit reserve. The t2.medium we're on now has 2 core and 24 CPU credits, which means that we can run both cores full throttle for 12 minutes an hour. A t2.large would give us 18 minutes an hour. An m3.medium would actually give us a little less RAM than our t2.medium, but would supply us 100% of 1 core. I'm not sure that's enough though, we'd probably need a m[n].large.
-
@accalia said in AWS issues:
I tried asking on IRC several times when ben_lubar was online and got no answer, i reached out to boomzilla as well, neither request got any answer. i mentioned both on these forums before and have
I thought I answered. I remember saying something about this. Which was that it was something that only Ben or Alex could possibly fix.
-
@boomzilla said in AWS issues:
@accalia said in AWS issues:
I tried asking on IRC several times when ben_lubar was online and got no answer, i reached out to boomzilla as well, neither request got any answer. i mentioned both on these forums before and have
I thought I answered. I remember saying something about this. Which was that it was something that only Ben or Alex could possibly fix.
which is all well and good, and i apreciate that, but it really is a non answer, it directed me towards the correct people but did nothing to provide me with an answer to my query.
particularly when there is effective radio silence from @ben_lubar and @apapadimoulis
-
@accalia said in AWS issues:
but it really is a non answer,
What other answer would you expect from anyone who isn't Ben or Alex?
-
@ben_lubar said in AWS issues:
@Yamikuronue said in AWS issues:
How much disk space have we eaten up in the past couple years of image uploads and rambling on?
{useful numbers}
Based on that, I could probably provide a VPS that I'm not using that would fill the needs.
-
@abarker said in AWS issues:
@accalia said in AWS issues:
but it really is a non answer,
What other answer would you expect from anyone who isn't Ben or Alex?
maybe help getting an answer from them given my reaching out to them was ineffective?!
that's why i asked for the help in the first place.
-
@accalia said in AWS issues:
that's why i asked for the help in the first place.
We have a category for that
Filed Under: Okay, I'll stop and go to sleep
-
Again, I'd like to be thankful that we avoided all those toxic hellstew forums
-
@bb36e said in AWS issues:
Again, I'd like to be thankful that we avoided all those toxic hellstew forums
Remember Community Server?
-
@ben_lubar CS does not qualify as a forum
-
Oh my, this is indeed slow....
AWS pricing is baffling. It's like, not just the instance size, but all the other bullshit like transfer fees, I/O fees, multiple-fee fees, etc.
IIRC, when we first set up the nodebb server, it was like $400/month or something, because of the IO1 disk or something. Then we went to small, now medium. The total AWS bill now is ~ $200/mo for this server, and a t1.micro that hosts http://omg2.thedailywtf.com/ -- I think we can try to upgrade a bit more, but it's starting to creep back up. I feel, at this price, we could get a dedicated quad core server that's better, faster, and probably similar price!
the main site is hosted at hivelocity: https://www.hivelocity.net/dedicated-servers/#products
our contract is renewing with them in December, so maybe we should just get a dedicated server there?
-
-
@sloosecannon said in AWS issues:
@apapadimoulis said in AWS issues:
$200/mo
o.O
Holy crap.
And yea, I think we can do better, especially if that's how much we're throwing at AWS right now. I mean, you can get this for ~half of that, with 8 cores and 128 gb of RAM. It doesn't have an SSD but you could probably run the freakin database in a ramdisk if you needed to!
-
HiVelocity's cheapest server What we have now CPU E3-1230 v5 3.4GHz Skylake E5-2676 v3 2.4GHz Haswell Cores 4 + hyperthreading 2, but you can only use 40% of a single core with an additional 40% that you can save up Memory 16GB DDR4 4GB (I can't find more information even with system tools) Disk 240GB SSD 200GB SSD OS Ubuntu 16.04 LTS Ubuntu 16.04 LTS Network 20TB/month, 1Gbps currently using 25GB/day, unlimited if you have unlimited money Uptime SLA 99.99% (52½ minutes) 99.95% (4⅓ hours) Price/month $129 $122.35
-
@ben_lubar No brainer. AWS is only sane for applications that scale and make financial sense to scale elastically instead of just buying for the peaks. Or when you're WtfCorp and your non-AWS datacenter vendor makes AWS look cheap.
Also, the AWS image is Ubanter? I thought it was some weird custom Debian derivative unless you sprang for SLES/RHEL
-
@e4tmyl33t said in AWS issues:
On-demand servers are more expensive than reserved instances, too, if you're not planning to lock in for a 1 or 3 year period.
Amazingly, Amazon emailed me telling me o was by not using reserved. Helpful.
-
@Weng said in AWS issues:
Ubanter
@Weng said in AWS issues:
some weird custom Debian derivative
Same thing, no?
-
@Weng said in AWS issues:
Looking at these adorable baby t2's, there is no fucking way I'd ever use them for production
though it's running what amounts to a static site ( status.tsaukpaetra.com )...
-
@ben_lubar said in AWS issues:
@CarrieVS said in AWS issues:
It seems like this needs sorting, but in the immediate term, would it help if us regular users all promised to stay off the forum for a day to let the CPU credits recover?
I'd much rather have the regulars using the forum than the search engine bots.
503 them with a try-again of three days?
-
@pydsigner said in AWS issues:
There is no way we should be paying 1400/year to host this forum. I mean, may as well go dedicated at that point.
I was wondering that as well while reading about all those AWS complications. I have a dedicated box at OVH that costs some €600/y. 8-core E3-1245 @ 3.4 GHz, 32 GB, 2 or 3 TB (I forget, never used all of it so far), 250 Mbit. Why would anyone pay several times that for an AWS instance if the software supports clustering so it doesn't depend on dynamic instance upscaling?
-
With all this talk of AWS and tiers-- maybe I'm just being dense or missing something but--
Why aren't we on the tier of "A fucking low/mid range Dell server in Indeo's offices", with the scaling/maintenance plan of "we can do whatever the fuck we need to".
Instead of getting locked out of an third-party server, we could-- you know-- not.
I mean, we're talking about dropping over a thousand bucks a year here. Maybe I'm missing some of the subtle nuances and hellishly complex engineering required in running a goddamn forum for less than 100 people... but I suspect not.
I would be absolutely shocked if @apapadimoulis didn't have the hardware needed sitting unused in a closet collecting dust or, like, being a shelf for scotch bottles.
-
-
@ben_lubar said in AWS issues:
@Lorne-Kates said in AWS issues:
less than 100 people
1000 hourly unique visitors
Again, less than 100 people. People who do things like-- post. And breath.
-
@Lorne-Kates said in AWS issues:
@ben_lubar said in AWS issues:
@Lorne-Kates said in AWS issues:
less than 100 people
1000 hourly unique visitors
Again, less than 100 people. People who do things like-- post. And breath.
We have 601 users with more than 100 posts. But if we use that logic to determine who a people is, @Remy isn't a people.
-
@ben_lubar said in AWS issues:
But if we use that logic to determine who a people is, @Remy isn't a people.
Correct.
But there has to be some sort of metric-- volume of posts by the top X active users-- that determines if a forum is "high bandwidth traffic" or not.
I am willing to bet our forum is way, way under any sane threshold of "busy". This isn't Something Awful or a chan or something.
-
@Lorne-Kates said in AWS issues:
high bandwidth traffic
Let me put it this way. We have 25GB of outgoing data per day. That might not be multiple terabytes, but I certainly can't upload 25GB in a day.
-
@ben_lubar said in AWS issues:
@Lorne-Kates said in AWS issues:
high bandwidth traffic
Let me put it this way. We have 25GB of outgoing data per day. That might not be multiple terabytes, but I certainly can't upload 25GB in a day.
I'll bet if the composer didn't round-trip for every keypress we could reduce that significantly...
-
@ben_lubar said in AWS issues:
@Lorne-Kates said in AWS issues:
high bandwidth traffic
Let me put it this way. We have 25GB of outgoing data per day. That might not be multiple terabytes, but I certainly can't upload 25GB in a day.
You aren't a good baseline. You couldn't upload 25MB per day. (BAM Milwaukee PC sickbern!)
How in the shit do we put out 25GB of traffic a day? I run a wordpress site for a mid-level Internet video maker with thousands of visitors. The site self hosts images. It puts out maybe 100GB a MONTH.
The more I hear about NodeBB's resource usage, the more baffled I get. It is absolutely mind-blowing to me that anyone can expect a forum to require a mid-range AWS that costs thousands per year just to function.
Either we have something configured wrong, or NodeBB is wrong right down to the very core.
-
@Lorne-Kates said in AWS issues:
Either we have something configured wrong, or NodeBB is wrong right down to the very core.
Sounds like the EMACS of the 10s: Eight Cores And Constantly Throttling
-
@Lorne-Kates said in AWS issues:
Either we have something configured wrong, or NodeBB is wrong right down to the very core.
$ head -n 1 /var/log/nginx/access.log | cut -d ' ' -f 4-5 [20/Oct/2016:06:25:23 +0000] $ wc -l /var/log/nginx/access.log 776959 /var/log/nginx/access.log
That's about 32kB per page, including images. Not too bad.
-
-
@LaoC said in AWS issues:
@Lorne-Kates said in AWS issues:
Either we have something configured wrong, or NodeBB is wrong right down to the very core.
Sounds like the EMACS of the 10s: Eight Cores And Constantly Throttling
That's an interesting way of spelling Cores.
-
@ben_lubar said in AWS issues:
watchdog.bash
@ben_lubar said in AWS issues:
watchdog.bash
I hope you realize that you are not running a life-critical embedded system that needs a watchdog for the occasional gamma rays from distant stars.
Are trying to live up to the forum name and be a good example? The watchdog problem was months ago, did you run under gdb to see what is it that segfaults? I bet it is the Go JIT thingy not playing well with PHP. Why? Why Go+PHP?!!!?
-
-
@ben_lubar said in AWS issues:
@dse said in AWS issues:
Why Go+PHP?!!!?
@dse said in AWS issues:
Go JIT
The original flame graph I remember you had Go extensions that were JIT compiled. Where as the majority of forum is in another language (PHP?, or JS?)
-
@Tsaukpaetra said in AWS issues:
@Weng said in AWS issues:
Looking at these adorable baby t2's, there is no fucking way I'd ever use them for production
though it's running what amounts to a static site ( status.tsaukpaetra.com )...
Also regarding that, should I be concerned if said instance randomly decides to spike its cpu usage for seven minutes or so all the sudden? I keep getting alerts mentioning how this is happening but um... No idea what the cause is and don't want to figure out where logs that might be useful reside...