#influxdb

I'll bump to 0.8.5 and see if it makes any difference :)

I am running into a crash consistently after about an hour of uptime on the influxdb server. The error I get is a runtime: program exceeds 10000-thread limit

fatal error: thread exhaustion

hmmm what are you pushing to hit that many threads?

nathanielc, anywho this is your kernel

nathanielc, cat /proc/sys/kernel/threads-max

nathanielc, if you seriously think you need more threads running feel free to up it

cat somevalue > /proc/sys/kernel/threads-max

replacesomevalue with a real value

cat /proc/sys/kernel/threads-max

191904

and that's a lot of threads

hope everything is working right

I read up on the go error and it seemed to imply that the 10,000 threads where in a blocking state.

nope

that's a kernel limit really

nathanielc, by default that's the kernel limit

not sure why you show 191904 there

k

yeah i haven't tweaked it

yeah i noticed one box is an even 10000

and another is 1030439

weird

anything else in your logs to indicate other errors?

no, in fact I had to run the server non-daemon just to get that err, it came out over stderr and was not captured in the log

I am on fedora 19 with 0.8.5 install of influxdb

It has happend at least 6 times now, every time I run the server after about an hour it crashes

I am htting it with 50 metrics a second with some continuous queries running every 2s 4s 8s and 24s

shouldn't be too much

how many clients pushing?

one

eh then that should be a non-issue

here is the stacktrace at least the first 13k http://pastebin.com/UskVcEig

can't speak for fedora but I run it with docker

kinda why I like go apps; it's not really complex

it should just work

yep

for what its worth while I have one client the client writes only a single metric per call and its from several different go routines

I'd say opening an issue would be helpful detailing as much information as you can

on it thanks