This message was deleted.
# support
s
This message was deleted.
s
What is the system resources usage (CPU and memory) when you run into this problem?
j
I haven't checked CPU, but I increased mem from the default 100m to 500m, that didn't make much of a difference - I assume because it's just buffering data until it can send it, but it never gets sent
s
Is is 500mb? That's too small -- at least give a 2GB.
j
Well, it was running with 100m just fine until I upgraded
Actually that's just what it requests - the usage is much higher, around 5-6GB when it's killed
Here's how it looks - the collectors are the ones going up...
Zoomed in on one - here it's also clear that there's a short period in the beginning where everything works, then it can't connect, and memory just goes up
s
Both issues have common underlying issues i.e resource usage. If you could share the heap profile of the collector process that would help us better understand the issue
j
But isn't it strange that the increase in memory usage coincides with the io error?
s
No, I don't think it's really that strange. The i/o error occurred because of the broken connection. The connection might have been terminated by ClickHouse under the system resource stress.
j
Isn't it then strange that it doesn't reconnect again?
s
It would reconnect. How did you confirm it didn't reconnect?
j
Well, it stops sending anything, then after a while it's OOM killed, then it's restarted and it works again. You see the pattern pretty clearly on the last screenshot I sent where the memory usage is flat initially, then it's interrupted and doesn't send anything until it's killed