Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Kiezer struggling #299

Closed
gravitystorm opened this issue May 7, 2019 · 11 comments
Closed

Kiezer struggling #299

gravitystorm opened this issue May 7, 2019 · 11 comments

Comments

@gravitystorm
Copy link
Collaborator

Similar to #284, tiles service here has been very slow today. It seems to be a problem with keizer

The network graph shows drops in the expected diurnal cycle of traffic:

if_eth0-day

The squid times shows median service times measured in multiple seconds

squid_times-day

The CPU chart shows the machine is pretty busy, and that something is stealing from the (presumably virtualised) CPUs.

cpu-day

@tomhughes
Copy link
Member

Sadly we don't have the resources to manage these issues - we have far more caches that we are reasonably able to monitor or manage and much of their performance is outside our control anyway.

@tomhughes
Copy link
Member

Also because keizer is just one quarter of the German cache we have no way to control the load going to it on an individual basis. I believe kalessin is also getting too much load but the same problem applies.

It's not really something we can solve without a complete redesign of the cache system...

@tomhughes
Copy link
Member

Sorry @gravitystorm I wasn't necessarily trying to say we should close this... I'm just a bit jaded trying to keep the cache show on the road and finding that every time you fix one machine it moves the problem somewhere else :-(

I'm a bit surprised Poland is on the German cache as I thought it pretty much only did Germany so we could probably move Poland off it.

Germany is always a problem due to the massive load and fact that the "cache" is actually four machines of rather different capabilities being round robined via DNS.

@tomhughes tomhughes reopened this May 7, 2019
@tomhughes
Copy link
Member

Hopefully openstreetmap/dns@cef3ef9 will help.

@gravitystorm
Copy link
Collaborator Author

Sorry @gravitystorm I wasn't necessarily trying to say we should close this... I'm just a bit jaded trying to keep the cache show on the road and finding that every time you fix one machine it moves the problem somewhere else :-(

Oh, I can understand the jadedness, I just felt that having this specific issue open was therefore not going to help solve any of the wider problems. If it was a specific problem with the specific server, then that would be different, but since it's just part of the wider tile serving situation then I was happy just to close it.

Thanks for looking into the situation though!

@AndreSchwarzer
Copy link

@tomhughes && @gravitystorm Would another server on the same or different provider help?

BTW I'm the sponsor of keizer 😄 👨‍💻

@tomhughes
Copy link
Member

Tricky... Germany is a bit of a nightmare, being split across four servers of differing capacities :-(

@AndreSchwarzer
Copy link

So a "clone" (same capacities) of keizer on cloud.hetzner.de too ?

So it'll be 5 servers of 4 settings :P

@Firefishy
Copy link
Member

@AndreSchwarzer yes, another server would be greatly appreciated. Best is via email operations AT osmfoundation. Org

@AndreSchwarzer
Copy link

I am resending my mail from August 08 2019

@AndreSchwarzer
Copy link

AndreSchwarzer commented Aug 28, 2019

You should have mail @Firefishy

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

4 participants