On 6 September 2017 at 12:30, Kevin Fenzi <kevin@xxxxxxxxx> wrote: > On 08/31/2017 04:54 PM, Stephen John Smoogen wrote: >> Proxy01 paged on a disk space issue this evening. Half of the disk >> space usage was journal logs. I vacuumed them down to 30 days which >> gave us about 2GB free. The other disk usage is koji.fedoraproject.org >> and kojipackages.fedoraproject.org which are using 1.8GB of disk space >> in the last 24 hours. >> >> The lion share of this log is >> >> 10.5.125.62 - - [31/Aug/2017:04:02:16 +0000] "POST /kojihub HTTP/1.1" >> 200 4118 "-" "koji/1" >> >> (with the POST being the size of whatever >> koji02.phx2.fedoraproject.org. is pushing to koji.fedoraproject.org >> (which somehow sounds weird).) > > Yeah, I am not sure how to fix this, but we have several scripts running > on koji02 (from koji-gc) that do a lot of queries and remove signed > copies and old builds that aren't refrenced anymore. In the past, it > just used localhost so the traffic was local, but now it uses > proxies->koji01 like everyone else. :( > > I think it might be worth making a pair of proxies: > proxy100 and proxy101 that handle only internal phx2 traffic. This would > allow us to not mess up internal stuff when there is heavy load on > proxy01/10. > > Thoughts? > This sounds like a good idea. proxy01/10 load can get really bogged with builds while also trying to pull stuff for external customers. > kevin > > > > > _______________________________________________ > infrastructure mailing list -- infrastructure@xxxxxxxxxxxxxxxxxxxxxxx > To unsubscribe send an email to infrastructure-leave@xxxxxxxxxxxxxxxxxxxxxxx > -- Stephen J Smoogen. _______________________________________________ infrastructure mailing list -- infrastructure@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe send an email to infrastructure-leave@xxxxxxxxxxxxxxxxxxxxxxx
