smoke: saturate a peer

[FintanH]

Trying to replicate a dead-lock that we witness in a scenario between a client and a seed node.

The real world scenario: The client periodically creates a test every few seconds. On the 12th project the seed does not replicate it. There's no output from logging nor GIT_TRACE_PACKET.

This test scenario: Peer 1 creates 12 projects and Peer 2 tries to replicate them all in a loop. The dead-locking is not witnessed and the test passes.

Signed-off-by: Fintan Halpenny fintan.halpenny@gmail.com

[kim]

I think you forgot to add a file there..

[FintanH]

I think you forgot to add a file there..

I did :grimacing: pushed now though :joy:

[FintanH]

Going to try to emulate peer 1 gossiping the new projects and peer 2 replicating when they hear the gossip

[kim]

Only question here: do we really gain anything by running the test? I'm not sure it is the case that it would catch the underlying issue if it was re-introduced as a regression. As it's also not a cheap test, it means that it just creates heat.

If it's valuable to have it as a diagnostic/forensic tool, perhaps it could go into e2e as an executable?

[xla]

Only question here: do we really gain anything by running the test? I'm not sure it is the case that it would catch the underlying issue if it was re-introduced as a regression. As it's also not a cheap test, it means that it just creates heat.

The smoke test does not complete since the number of projects was upped, which seems to force to hit the upper limit on the streams. I'd still consider it an accurate gauge if this problem is addressed or not, it doesn't hover show the relationship properly which is communicates poorly and will likely bamboozle new and current maintainers in six months alike.