NFS shares using a cache disk don't properly spin up data disks (rc 16c)

[guitarlp]

I'm new to NFS but I found a bug regarding NFS shares and cache disks.

 

I had originally setup an NFS share (set to use the cache disk) for my ESXi automated backups using ghettoVCB. Essentially, ESXi kept the NFS share mounted, and a script on the server ran through a cron job to take snapshots of each VM and copy them to the mounted NFS volume. I can the script, testing through a cron job, and saw that everything was working well.

 

Fast forward a couple of days and I checked the backup location on unRAID to see that nothing new had been copied since my original backups. I SSH'd onto the ESXi machine and navigated to the NFS share mount directory. An ls command showed nothing in that directory. When I tried to run the backup manually I got errors saying the directory didn't exist. Once I spun up all disks, my ls command worked and when I ran a backup from within ESXi it processed successfully.

 

I originally assumed this was a fault of NFS, but someone on here suggested not using the cache drive. Sure enough, if I disable the cache drive and all the disks are asleep, ESXi is able to spin up the drive through the NFS share. So for now, for shares that use NFS I have to disable the cache drive. Otherwise, the disks for those shares won't spin up and my NFS mounts won't function correctly.

[limetech]

I cannot duplicate this (i.e., it works for me).  Maybe your client is timing-out waiting for the drive to spin up?

 

Can anyone else reproduce similar behavior?

[guitarlp]

I'm not sure. But I know that this has been an issue for me on the following machines:

 

Macbook running 10.8.4

Ubuntu Server 12.04

VMware ESXi

 

Since this isn't limited to my one machine, I would assume it's something on unRAID's side. That being said, I also have an issue with using Time Machine on unRAID and having all my disks spin up, and no one else seems to be having that issue. Sounds like some sort of config issue on my end, but I did wipe my 4.7 flash and start clean with 5.0 rc16c.

[guitarlp]

It's also worth noting I'm running unRAID within ESXi, other's that don't have my issues are likely running bare metal. So it's possible that the ESXi layer is causing this issue. I'll try doing some testing booting directly into unRAID bare metal and see if that does anything different for me.

[limetech]

It's also worth noting I'm running unRAID within ESXi, other's that don't have my issues are likely running bare metal. So it's possible that the ESXi layer is causing this issue. I'll try doing some testing booting directly into unRAID bare metal and see if that does anything different for me.

That would be a  big help, thank you.

[guitarlp]

I switched my machines from using fstab to mount the NFS mounts, to instead using autofs. So far, unRAID has been spinning up disks without any issues (even within ESXi). I think it's safe to close out this issue for now. If I run into issues again I'll do some more testing and report back my findings in this thread.

[newsboy]

I've been having problems with NFS.  When I was mounting the user shares through fstab I would get stale NFS data errors. Upon the suggestion in this thread I gave autofs a try. I'm still running into problems, but I now intermittently get an error saying a directory does exist. Restarting autofs seems to clear things up, but by that time my plex box, which runs on the same host but different VM, has decided the directory with the TV shows in it no longer exist.

 

I'm running unraid 5.0 final release. ESXi 5.5, IBM M1015 HBA, and I have a cache drive.

 

I'm thinking about pulling out the cache drive for now. I know having or not having the cache drive made no difference when I was mounting NFS through fstab.

 

Suggestions?