Running the latest RC 10b.

I have multiple dockers running (Plex, nzbget, sonarr, radarr, etc)

Dual SSD running as a cache pool (Samsung Evo 840 250GB, Evo 850 500GB)

Symptom: All plex streams freeze, all docker container's webpages inaccessible.

Cause: When nzbget finishes download and starts moving large movies (15GB+) from /mnt/user/downloads/dst/Movies/ to /mst/user/all_media/movies the server freezes until transfer is complete.

Plex uses cache drive as temp transcode folder (/mnt/cache/plexTranscode), nzbget uses download destination (/mnt/user/downloads/), plex media folders are located (/mnt/user/all_media/)

 

I could use some help in identifying how to remedy this situation before the wife divorces me for interrupting her shows.

Isn't it just copying from one folder to another on the cache drive during this transfer period? Should I turn off "use Hardlinks instead of Copy" from within Radarr?

I have the same issue, along with other users. Have a look at this thread:

 

 

So far as I remember, the only observation made thus far was that a single XFS cache device does not exhibit this behavior. Not useful for those of us who want redundancy on our cache - anyone who uses VMs or Docker, really - but an interesting data point.

 

Has @limetech been looking at this issue at all recently? Any other tid-bits that might help the community narrow down the problem any?

You might try using a download destination that is not parity protected. I had similar issues with deluge and using a UD mounted drive to download to fixed the problems for me.

Read through the other forum post and trying page 2 @johnnie.black 's

btrfs balance start --full-balance /mnt/cache

to see if this will clean it up.

 

Total devices 2 FS bytes used 46.86GiB
	devid    1 size 232.89GiB used 222.91GiB path /dev/sdk1
	devid    2 size 465.76GiB used 222.91GiB path /dev/sdj1

devid 1 (if I'm using the terminology correctly) had a whole lot of slack space that was not being trimmed.

 

I am running the command in putty... is it normal for it to just sit there after running the command with no feedback for awhile? It just seems to have put in a carriage return and there is no status update for the 5 minutes it has been running.

Edited November 10, 20178 yr by blurb2m
Adding more info/question.

It's running!

	Total devices 2 FS bytes used 45.97GiB
	devid    1 size 232.89GiB used 61.97GiB path /dev/sdk1
	devid    2 size 465.76GiB used 61.97GiB path /dev/sdj1

Balance on '/mnt/cache' is running
178 out of about 237 chunks balanced (179 considered),  25% left

1 minute ago, blurb2m said:

devid 1 (if I'm using the terminology correctly) had a whole lot of slack space that was not being trimmed.

 

Yes, there was a lot of slack on both, and this can result on most free space remaining untrimmed.

@johnnie.black Is there a way to schedule this full balance of the cache pool weekly?

If run weekly you don't need a full balance, this is what I do, it will be faster and less wear on the SSD(s):

 

btrfs balance start -dusage=75 /mnt/cache

 

You can use the user scripts plugin to schedule it.

6 hours ago, johnnie.black said:

If run weekly you don't need a full balance, this is what I do, it will be faster and less wear on the SSD(s):

 

btrfs balance start -dusage=75 /mnt/cache

 

You can use the user scripts plugin to schedule it.

 

@johnnie.black, is there a benefit to using this as a user script vs using the Dynamix SSD trim plugin?

8 hours ago, wreave said:

 

@johnnie.black, is there a benefit to using this as a user script vs using the Dynamix SSD trim plugin?

 

This should be used in conjunction with the trim plugin, they do different things, balance is to keep the allocated space under control, because currently most of the empty allocated space won't be trimmed, and if left uncheck for some time it can lead to ENOSPC errors despite having lots of free space.

On 11/10/2017 at 11:57 AM, blurb2m said:

It's running!

 

	Total devices 2 FS bytes used 45.97GiB
	devid    1 size 232.89GiB used 61.97GiB path /dev/sdk1
	devid    2 size 465.76GiB used 61.97GiB path /dev/sdj1

Balance on '/mnt/cache' is running
178 out of about 237 chunks balanced (179 considered),  25% left

Is the Total devices display gone in the latest 2 RCs?