mk_server: race condition for `MK_SERVER_SIGNAL_START` between worker and main threads about monkey HOT 5 CLOSED

@edsiper Is this something we could bring into Fluent Bit 1.9?

from monkey.

hi @qingling128 , if this has been extensively tested in G side, yeah, definitely, we can schedule another 1.9 release.

from monkey.

Hi @edsiper, I have done testing of this by using this standalone test program and by patching it into Fluent Bit and using it in the Ops Agent, which uses a config that starts a monkey server using the Prometheus exporter (see our default config here). I'm happy to test any other scenarios that might cover other use cases not covered by this.

from monkey.

Hi @edsiper We’ve tested the custom build on GCE Windows 2012 and 2019 VMs with 10+ restarts, and the fix worked fine. We were able to reproduce the problem without the fix. So far, it looks good. What additional testing do you think is preferred? Is this sufficient?

from monkey.

@qingling128 thanks for the clarification. would you please open a github issue in fluent/fluent-bit asking to upgrade lib/monkey ? (we handle the PR for that)

from monkey.