On Thu, Oct 11, 2012 at 6:23 AM, Steve Dickson <SteveD@xxxxxxxxxx> wrote: >>>>>> Per Aníbal Salazar, I'm sending this to the nfs mailing list.. >>>>>> >>>>>> ========== >>>>>> >>>>>> Hi. I would like to know who I can talk to about having the rpcbind's >>>>>> timeout value settable on the command line by the user. In many cases >>>>>> the timeout is too long, requiring hackish solutions. It would be >>>>>> best, and makes sense, that the user should be able to set the timeout >>>>>> to something other than the default value if he chooses. If you could >>>>>> direct me to the right person to talk to about it, I'd appreciate it. >>>>> >>>>> What timeout are you referring to? The one given to poll()? >>>> >>>> Hi. I guess so but not really sure. I'm talking about the timeout that >>>> happens when rpcbind is waiting for a response. Sounds like poll() >>>> could be it. We have an nfs server on .100 and the response happens >>>> immediately. >>>> >>>> $ rpcinfo -t 192.168.1.100 nfs >>>> program 100003 version 2 ready and waiting >>>> program 100003 version 3 ready and waiting >>>> program 100003 version 4 ready and waiting >>>> >>>> but there's no server on say .101 so if we run the same command on >>>> that ip, the timeout takes a very long time. It's this timeout that >>>> should be user-definable on the command line in my opinion. Any >>>> thoughts about it? >>> >>> Hmm... I'm guess that is the 7min tcp connect time out cause by >>> the -t option... Try using -u instead of -t... Basically using >>> UDP instead of TCP... In general I would never recommend that >>> but in this particular case it might help... >> >> Thanks for this suggestion. I tried with -u but the timeout still >> takes at least 1 min. Is it not feasible to have a command line >> timeout where users can set it to something appropriate for their >> needs? For example, in our case we only need about 5 seconds at most. > > hmm... when I do a "rpcinfo -t <ip-address> nfs" to a machine that > does not have a daemon listening I immediately get: > rpcinfo: RPC: Port mapper failure - Unable to receive: errno 111 (Connection refused) program 100003 is not available > > So I not seeing here this hang is coming from... The computers on our network run a mixture of different OS'es so maybe that is relevant. Regardless though it makes sense that we should be able to tell rpcbind to abort if I hasn't received a response within X seconds. That's much better than being forced to wait predefined timeouts, or timeouts in other places. Or is it just me? Cheers -- To unsubscribe from this list: send the line "unsubscribe linux-nfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
