Concurrent map read and map write when multiple reads

[cznic]

Cf.: weaveworks/flux#393.

[gernest]

I wonder if there is away to reproduce this. It looks nasty.

[cznic]

Unfortunately, we have no answer from @philwinder yet.

[twitchyliquid64]

I encountered this with multiple writes and a single read. Was unable to reproduce again, despite doing the same thing for the next two weeks. There were over 200 go-routines running at the time of the crash.

If we can't repro, can we put in some recover() at the top of ql somewhere, and collect information (that can be dumped to the console, before rebubbling the panic) that will help debug in the future?

[cznic]

If we can't repro, can we put in some recover() at the top of ql somewhere, and collect information (that can be dumped to the console, before rebubbling the panic) that will help debug in the future?

Provided the error is recoverable, which I don't know, I like the idea. I'm sending you a collaborator's invitation in case you would like to create the patch. If so, please make it a separate branch for easy reviewing before merging it eventually, thank you.