While working with larger repositories (.git/ folder is 778 MiB in my case), I find my public node frequently emits warnings like
WARN worker Read channel timed out for upload-pack rad:<rid>
I’d like to suggest to make the default fetch timeout longer to be 300s (5 minutes), which is more balanced than the previous 1 hour or 3 seconds timeout.
In the future it might also be desirable to allow configurable fetch timeouts, but this patch does not address this.
Signed-off-by: Gaoyang Zhang gy@blurgy.xyz
While working with larger repositories (.git/ folder is 778 MiB in my case), I find my public node frequently emits warnings like
WARN worker Read channel timed out for upload-pack rad:<rid>
I’d like to suggest to make the default fetch timeout longer to be 300s (5 minutes), which is more balanced than the previous 1 hour or 3 seconds timeout.
In the future it might also be desirable to allow configurable fetch timeouts, but this patch does not address this.
Signed-off-by: Gaoyang Zhang gy@blurgy.xyz
Thanks for the patch, and sorry for the VERY delayed reply.
We went back and forth on this timeout value for a while originally. After getting a chance to actually investigate the issue there are now a series of patches that will fix this:
- f48c03698bcb2c015829f0c8e04240b95f3a7fca
- b6b90eaaf22180bb5235de1d212ccf54eed35681
- 49513bf467a0d3e8a41283edb7faf40ad52b4354
- e88429499d64b02b833e80d6e702854dcbfb46c5
Once these are merged, this timeout issue will go away – we’ve tested it on repositories up to 8GB in size.
So, all this said, I’m going to archive your patch in favour of the above :)