Hi Jeremy,
patchwork.ozlabs.org has been slow for us for some weeks, in a browser
and through the REST API, and requests often time out. I measured it
today: between 10:58 and 11:40 UTC I fetched three U-Boot URLs 30 times
each with curl and a 60 s limit, from a host in Europe, and one
patchwork.kernel.org page from the same host for comparison:
median p90 max >10 s timeouts
/project/uboot/list/ 4.7 s 18.1 s 37.5 s 9/30 2/30
/api/1.2/patches/?project=uboot 4.7 s 17.0 s 34.3 s 9/30 2/30
/api/1.2/patches/2294003/ 4.7 s 18.0 s 35.5 s 6/30 2/30
patchwork.kernel.org list page 0.9 s 0.9 s 4.9 s 0/30 0/30
The TCP connect takes 0.3 s and the rest is the wait for the first
byte, so the time goes in the server. The six timeouts came in two
groups, 11:01-11:02 and 11:16-11:21 UTC, where every URL failed.
Is this a known load problem on the machine, and is there anything
users can do to lighten it, for example in how often they poll the API?
I can send the raw timings if they help.
Thanks,
Mehmet
