LinHES Forums
http://forum.linhes.org/

7.1 diskless frontend install
http://forum.linhes.org/viewtopic.php?f=22&t=22393		 Page 1 of 1
Author:  poflynn [ Tue Jan 03, 2012 12:37 am ]
Post subject:  7.1 diskless frontend install
Hi all,

I'm getting 2 of these up and running, for the most part, so far so good. I do have one annoyance however. When I boot up the diskless FE it sits at this screen for a good 5 minutes at least, and then proceeds as normal. There's nothing (that I can see) in /var/log/messages which isn't surprising as root isn't mounted yet. I don't believe this was even a step when booting diskless in R6 (root was mounted before you got to these white text msgs).

As it ends up booting ok, this is not a show stopper but it definitely adds to the boot up time! Any ideas?

Thx,

Paul

Image
Author:  poflynn [ Fri Jan 20, 2012 10:55 pm ]
Post subject: 
Just to follow-up a little: The only entries in the backend log are below. tftpd seems to be on GMT?? Ignoring that for now.. The nfs mount request entry happens as soon as the FE is done loading the kernel. There are no other entries, implying the FE is not successfully mounting root after the ~4min delay but before the delay. This is backed up by the timestamps in the two logs below (the clocks are the same thx to ntp).

Any ideas?

Thx.

Backend log:
Quote:
Jan 21 04:46:26 linhes7 in.tftpd[7068]: tftp: client does not accept options
Jan 20 20:46:41 linhes7 mountd[794]: authenticated mount request from 10.1.2.21:832 for /nfsroot/fe-1 (/nfsroot/fe-1)


Frontend log:

Quote:
Jan 20 20:50:56 fe-1 kernel: [ 0.000000] Initializing cgroup subsys cpuset
Jan 20 20:50:56 fe-1 kernel: [ 0.000000] Initializing cgroup subsys cpu
Jan 20 20:50:56 fe-1 kernel: [ 0.000000] Linux version 2.6.39-LinHES (root@dev) (gcc version 4.5.1 (GCC) ) #1 SMP PREEMPT Fri Oct 7 04:50:03 UTC 2011
Jan 20 20:50:56 fe-1 kernel: [ 0.000000] BIOS-provided physical RAM map:
Author:  poflynn [ Sun Jan 22, 2012 8:09 pm ]
Post subject: 
Made a little bit of progress on this one. Well, ok, a friend of mine did. It seems some nfs dependencies aren't yet running when the FEs try to mount stuff, see below. Note that this will be a problem for anyone attempting to run a diskless FE.

This is a problem because the NFS mounts (pacman cache & root) on the FEs happen right at the start in /etc/runit/1 which calls /etc/rc.sysinit which runs "mount -a -t nfs" and this is where it hangs. However rpcbind and nfs-common scripts are required for nfs but are not run until later, in /etc/runit/2 . Check out see the bit at the end, "runsvdir -P /service". Thus the above mount attempts hang, waiting for rpcbind & nfs-common.

Thus I think that "mount -a -t nfs" has no business being in that rc.sysinit script. It's too early.

I could remove the nfs mount command from rc.sysinit, but the question is, where to put it instead? It needs to run before mythbackend tries to start up in runit/2 . I have been looking at how runsvdir starts the things in /service but I can't figure out how it knows what order to run them in. Any idea?? If I knew for sure that mythbackend will run *after* the rpcbind and nfs-common, then I could just put the nfs mount command into the start of mythbackend itself.
Author:  bigtoedsloth [ Mon Jan 23, 2012 3:50 am ]
Post subject: 
poflynn wrote:
I have been looking at how runsvdir starts the things in /service but I can't figure out how it knows what order to run them in. Any idea??


I'm reasonably sure that there is no specifc ordering. But you could explicitly do an "sv up rpcbind" and "sv up nfs-common" early on in the mythbackend run script to force the order.
Page 1 of 1 All times are UTC - 6 hours
Powered by phpBB® Forum Software © phpBB Group
http://www.phpbb.com/