I got the final parts for my server on Friday including a 2nd cache drive to create a cache pool. 

 

 

What's confusing me is that some of my shares which I've now set to 'cache only' are alternating between green 'all files protected' and orange 'some or all files unprotected'.  What's doubly confusing is some shares like 'molly' are virtually empty and have had no new writes.  Why is this?  I would expect all cache only shares to go green and stay green, with any new writes made to both drives

 

 

Secondly, on the Main tab my two cache drives have different write stats - shouldn't they be the same?  I can understand that the reads might be different, but shouldn't the writes be the same once they are balanced?

 

 

Thirdly, out of interest - when multiple files are read from the cache does unRAID get File A from Drive 1 and File B from Drive 2 to maximise speed?

 

 

Thanks in advance.

dzmm-diagnostics-20161024-2042.zip

Secondly, on the Main tab my two cache drives have different write stats - shouldn't they be the same?  I can understand that the reads might be different, but shouldn't the writes be the same once they are balanced?

 

Click on cache to access the cache web page and post contents of "btrfs filesystem show"

Click on cache to access the cache web page and post contents of "btrfs filesystem show"

Label: none  uuid: 2da0c274-d64f-4947-a87c-bf52278a3293
Total devices 2 FS bytes used 64.70GiB
devid    1 size 232.89GiB used 109.00GiB path /dev/sdg1
devid    2 size 232.89GiB used 111.00GiB path /dev/sdf1

Looks normal

Looks normal

 

 

  Shouldn't both drives have the same amount of storage used?

 

 

I saw the 'balance' button for the first time on the cache page.  The first time it got to 12% left and then my server crashed (a lot going on, so not sure if this was the cause although this is probably only my 2nd crash).  On reboot both drives were showing same amount used although not all cache only shares were green (more than before though).  The 2nd balance run completed but still some cache only shares aren't green.

 

 

This doesn't make sense to me - I thought the whole point of the cache pool was so to provide redundancy?  What am I doing wrong?

That and crashing during a balance is not so normal, what does it show on "btrfs filesystem df"?

 

 

 

btrfs filesystem df:

Data, RAID1: total=78.00GiB, used=75.75GiB

System, RAID1: total=32.00MiB, used=16.00KiB

Metadata, RAID1: total=1.00GiB, used=183.14MiB

GlobalReserve, single: total=123.45MiB, used=0.00B

Also looks normal, if you're still having issues maybe delete cache pool and start over.

I just tried removing second drive, but my cache wasn't recognised with one drive so I had to put it back.

 

 

I'm running balance again to see if I can fix the two cache only shares that are orange triangles rather than green all protected circles.  Keen to get this sorted as it defeats the point of the pool if it doesn't have redundancy!

I just tried removing second drive, but my cache wasn't recognized with one drive so I had to put it back.

 

That is normal if you just unassigned one of the cache disks an started the array.

 

Staring over and reformatting both cache disks would probably be better, but if you want to before try without formatting remove one of the disks following this procedure:

 

http://lime-technology.com/forum/index.php?topic=48508.msg484479#msg484479

 

Then re-add it and see if issue is fixed, if not, format both and start over.

getting weirder and weirder...

 

I tried moving files from mnt/cache to the array so I could start with a fresh cache pool, but after I moved all files, on the Main tab it's still reporting 18.4GB usage even though there are no files showing in MC.

 

 

update: reboot solved.  Think had a file upload via rclone going on - rebuilding cache pool now

dzmm-diagnostics-20161028-0936.zip

ok, this is bonkers. 

 

 

I've worked out what the problem is, but I'll need help from the LT team to help me fix it.  Basically I can't create any cache-only shares that fall alphabetically between my shares /media/ and /plex_sync/ i.e medh works but not medib and plexa_sync works but not not plew_sync !!!!

 

 

My media share is a bog standard one that contains my photos, videos etc and is set to 'yes' for use cache.  It is my biggest share by far with over 10TB of files in.

 

 

My plex_sync is not standard.  I am currently syncing some of my Plex library to Cloud Sync and I'm storing the sync files while they are queued for upload in the plex_sync share.  Background sync files (cloud sync or device sync) are always stored in appdata, unlike the foreground sync files (live transcodes) which can be moved to another folder via plex's settings. Because I don't have space for hundreds of files in my appdata share I created a bind mount to move the relevant folder from my appdata to an the plex_sync array share at boot:

 

 

mount --bind /mnt/user/plex_sync/Transcode "/mnt/disks/sl308/appdata/plex/Library/Application Support/Plex Media Server/Cache/Transcode"

 

 

To get all my cache only shares to go green, I have to remove ANY /media/ files off the cache drive - even if I put just one 4KB file on the cache drive, unRAID will report anything between 'media' and 'plex_sync' as orange.

 

 

Maybe the problem is that because so many media files are on the array, the check to see if the cache pool is working properly fails and never completes.

 

 

Or, the bind mount is causing the problem but I doubt it (which is pointing to an unassigned drive), as removing media from my cache pool solves the problem.

ok, this is bonkers. 

 

 

I've worked out what the problem is, but I'll need help from the LT team to help me fix it.  Basically I can't create any cache-only shares that fall alphabetically between my shares /media/ and /plex_sync/ i.e medh works but not medib and plexa_sync works but not not plew_sync !!!!

 

 

My media share is a bog standard one that contains my photos, videos etc and is set to 'yes' for use cache.  It is my biggest share by far with over 10TB of files in.

 

 

My plex_sync is not standard.  I am currently syncing some of my Plex library to Cloud Sync and I'm storing the sync files while they are queued for upload in the plex_sync share.  Background sync files (cloud sync or device sync) are always stored in appdata, unlike the foreground sync files (live transcodes) which can be moved to another folder via plex's settings. Because I don't have space for hundreds of files in my appdata share I created a bind mount to move the relevant folder from my appdata to an the plex_sync array share at boot:

 

 

mount --bind /mnt/user/plex_sync/Transcode "/mnt/disks/sl308/appdata/plex/Library/Application Support/Plex Media Server/Cache/Transcode"

 

 

To get all my cache only shares to go green, I have to remove ANY /media/ files off the cache drive - even if I put just one 4KB file on the cache drive, unRAID will report anything between 'media' and 'plex_sync' as orange.

 

 

Maybe the problem is that because so many media files are on the array, the check to see if the cache pool is working properly fails and never completes.

 

 

Or, the bind mount is causing the problem but I doubt it (which is pointing to an unassigned drive), as removing media from my cache pool solves the problem.

 

If you don't have the bind mount in place, are there still problems?  That you can't create cache-only shares based on share name indicates to me, probably cache file system corruption.

 

From syslog it looks like you don't have any parity disks, in this case if a share directory in a 'cache-only' share does exist on the array, the icon next to the share name will be 'orange'.  Here's how it works.  First, if the array has valid parity we say that all array disks are "protected"; if no parity, they are "unprotected".  Second, if you have two or more cache disks, we say the cache is "protected", if only one cache disk it's "unprotected".  [Note if you manually have set multi-device btrfs cache pool to 'raid0' unRAID at present will not detect this and still consider cache "protected.]

 

The colored status indicator next to a user share indicates if that share exists on protected or unprotected storage.  That is, if the only place a share directory exists is on the array, and the array is protected, then the color is green.  Orange would indicate that the share directory exists on at least one unprotected storage set.

 

Remember the cache settings for a share only specify the possible target for new files created:

no - will try to create on array, if no space: error

yes - will try to create on cache, if no space create on array, if no space on array: error

only - will try to create on cache, if no space: error

prefer - (same as yes)

 

When a share directory listings are generated, cache (if present) and all mounted disks are searched for that directory, no matter what 'Use cache' setting.

 

Finally, btrfs "raid-1" is not true raid-1, in that they are not identical mirrors.  They contain the same data but it's possible the sizes reported in use for each device could vary slightly.

Thanks for all the info.

 

 

Sorry for the confusion, I CAN create cache only shares that fall alphabetically between 'media' (cache-only) and 'plex_sync' (array only but is linked via bind mount to appdata folder that sits on an unassigned SSD).  The problem is that these cache only shares show as orange even though all files be on the cache pool.

 

 

I think it is the bind mount that's creating the problems with the lights being orange rather than green.  I've just added my parity drive, so for peace of mind I think I might not create any cache only shares in that alphabetical range just so that all the folders go green.