Fwd: nfs kernel booting is slow

[Lloyd Parkes <lloyd%must-have-coffee.gen.nz@localhost>]

This mail got dinged by the mailing list HTML filter because Apple Mail switched the message format to HTML in the middle of the conversation.  

> Begin forwarded message:
> 
> From: Lloyd Parkes <lloyd%must-have-coffee.gen.nz@localhost>
> Subject: Re: nfs kernel booting is slow
> Date: 20 November 2017 at 2:50:15 PM GMT-10
> To: tech-net%NetBSD.org@localhost
> 
> 
> 
>> On 19/11/2017, at 1:03 PM, scole_mail <scole_mail%gmx.com@localhost> wrote:
>> 
>> Not Working:
>> 08:32:03.148935 IP 10.0.0.3.1020 > 10.0.0.1.2049: NFS request xid 41 104 read fh 0,0/-293204188 1024 bytes @ 30720
>> 08:32:03.148962 IP 10.0.0.1.2049 > 10.0.0.3.1020: NFS reply xid 41 reply ok 1124 read
> 
> xid 41 failed, response time 27μs
> 
>> 08:32:05.202903 IP 10.0.0.3.1020 > 10.0.0.1.2049: NFS request xid 41 104 read fh 0,0/-293204188 1024 bytes @ 30720
>> 08:32:05.202939 IP 10.0.0.1.2049 > 10.0.0.3.1020: NFS reply xid 41 reply ok 1124 read
> 
> xid 41 succeeded, response time 36μs
> 
>> 08:32:05.281127 IP 10.0.0.3.1020 > 10.0.0.1.2049: NFS request xid 42 104 read fh 0,0/-293204188 1024 bytes @ 31744
>> 08:32:05.281154 IP 10.0.0.1.2049 > 10.0.0.3.1020: NFS reply xid 42 reply ok 1124 read
> 
> xid 42 failed, response time 27μs
> 
>> 08:32:07.335172 IP 10.0.0.3.1020 > 10.0.0.1.2049: NFS request xid 42 104 read fh 0,0/-293204188 1024 bytes @ 31744
>> 08:32:07.335203 IP 10.0.0.1.2049 > 10.0.0.3.1020: NFS reply xid 42 reply ok 1124 read
> 
> xid 42 succeeded, response time 31μs
> 
>> 08:32:07.415459 IP 10.0.0.3.1020 > 10.0.0.1.2049: NFS request xid 43 104 read fh 0,0/-293204188 1024 bytes @ 32768
>> 08:32:07.415487 IP 10.0.0.1.2049 > 10.0.0.3.1020: NFS reply xid 43 reply ok 1124 read
> 
> xid 43 failed, response time 28μs
> 
>> 08:32:09.469354 IP 10.0.0.3.1020 > 10.0.0.1.2049: NFS request xid 43 104 read fh 0,0/-293204188 1024 bytes @ 32768
>> 08:32:09.469393 IP 10.0.0.1.2049 > 10.0.0.3.1020: NFS reply xid 43 reply ok 1124 read
> 
> xid 43 succeeded?, response time 39μs
> 
> It’s a really small sample set, but the failures are all <= 28μs and the successes are all >= 31μs. 
> 
> I’ve just seen your new working dump, which looks much better. Also, times in the working dump are shorter than the failed reply times in the not-working dump. The 100Mb/s link to the working switch will introduce a packet forwarding delay though, which might be helping.
> 
> Maybe it’s your new switch which doesn’t handle 10base-T very well?  Does the powermac do 10base-T full-duplex? Maybe the switch port has a duplex mismatch with your powermac? 
> 
> If it is a speed related issue, slowing down the NFS server’s network link should show this up. 
> 
> Cheers,
> Lloyd
>