Greyhole problems created because Disk Wizard does not mount the newly added partition
Greyhole.log may show one or more of the following messages:
WARN write: No metadata files could be created. Will wait until metadata files can be created to work on this file.
CRITICAL daemon: ERROR: The following storage pool drives are on the same partition: /var/hda/files/drives/xxxxx/gh, /var/hda/files/drives/yyyyy/gh, .... The Greyhole daemon will now stop.
WARN daemon: PHP Warning : mkdir(): No such file or directory in /usr/bin/greyhole on line 6235; BT: greyhole[L3353] run() => greyhole[L3322] initialize() => greyhole[L3340] set_metastore_backup() => greyhole[L6235] mkdir(/var/hda/files/drives/yyyyy/gh/.gh_metastore_backup)
You may also see Dashboard->Setup->Settings->Servers->Greyhole UI Server as stopped.
In the original forum thread I created on this issue, I thought the problem was that the Greyhole UI App was not creating the /gh subfolder. It is creating it, but it's "disappearing" later. Here's what's happening:
- User chooses the option to have the partition mounted automatically.
- Disk Wizard creates an entry in /etc/fstab for the newly added partition.
- Even though the user chooses the option to have it mounted automatically, it doesn't get mounted.
- User switches from Disk Wizard to Drive Pooling->Partitions and selects the newly added partition.
- Greyhole creates the /gh subfolder on the partition.
- If the user reboots or does a
mount -aor manually mounts the new drive, the mount point in /etc/fstab points to the full path that does not include /gh, resulting in the /gh folder seeming to suddenly "disappear." (Note that the reboot/mount will stop the error, "The following storage pool drives are on the same partition....")
- User begins copying files to the share such that the newly added partition gets a copy. Greyhole tries to create the necessary metadata file but the /gh folder doesn't exist at the mount point.
A temporary fix is to do the following under a sudo shell:
for i in /var/hda/files/drives/* ; do mkdir -p $i/gh ; done
Please consider making the following two changes to address this issue:
- In Disk Wizard, if the user elects to have the partition mounted automatically, mount it right away.
- When the user navigates to Dashboard->Drive Pooling->Partitions, a) do not allow the partition to be selected if the specified location is not a mount point, and b) give some sort of indicator on the Partitions tab that there's a problem with the partition. (One of the problems right now is that there's no clear visibility for the user that this issue exists.) In case it helps, in the code change for Issue #611 you'll find a similar check in
I can't seem to add screen captures as I write this on my Android tablet so will try shortly on my Windows desktop to attach them.
Also available in: Atom