Re: [Nbd] [PATCH] nbd: improve request timeouts handling

[Paul Clements <paul.clements@...856...>]

Michal,

thanks for this...a couple of preliminary questions...

On Mon, Jun 17, 2013 at 4:10 PM, Michal Belczyk <belczyk@...1274...> wrote:

The main idea behind it is to be able to quickly detect broken replica
and switch over to another when used with any sort of mirror type device
built on top of any number of nbd devices.

I'm wondering if a simpler change would have the same result. Namely, could you just apply the xmit_timeout, as currently implemented, to the recvmsg path (ignoring the timeout, of course, when no requests are queued). Is there a case that this would not handle that the proposed patch handles? They seem to be equivalent.

Also move nbd->pid modifications behind nbd->tx_lock wherever possible
to avoid races between the concurrent nbd-client invocations.

Is there a bug related to this that you've seen, or is this just a precaution?

 

Thanks,

Paul