On 12/30/2012 08:07 PM, Ian Kent wrote:
Yeah, that could be changed to exec showmount but the RPC code used for this is the same that's used to probe availability (essentially a NULL proc ping). The availability probe is needed all the time now now because, with mount.nfs letting the kernel do most of the work, there can be significant delays if a server isn't available. Unfortunately the RPC code needs to be fairly low level as well to control timeouts to give reasonable interactive response.
Well, arguably the right thing to do is the give the kernel *all* the information and let it sort things out. This is especially so since the kernel is the only part of the system which can possibly do a hot failover while mounted.
So this does need to be sorted out. I was thinking that it might be the error handling in the RPC code bailing out to early and not trying all addresses. Is the mountd server service also available and working over IPv6?
Yes.
For simple indirect (and direct) mounts you can prevent the probe being done by setting MOUNT_WAIT to a sensible value, say between 15 and 30 seconds, depending on your preference. Checking that will eliminate much of the parsing code as part of the problem, although it does sound like this is the RPC code anyway.
Probably. -hpa -- H. Peter Anvin, Intel Open Source Technology Center I work for Intel. I don't speak on their behalf. -- To unsubscribe from this list: send the line "unsubscribe autofs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
