ReadyNas 6.2 (NAS locks up entirely)

In my case on 6.1.9 never crash, after update to 6.2.0 beta NAS freeze after 1 or 2 hours, on 6.2.0 finall 1 freeze on week. I f only buy hdd i will backup data and roll back to 6.1.9. (you need change header of update file and facory reset you NAS).

My 316 locked up completely this evening in a very similar way.

Currently running the 6.2.1 proper release.

As with others, all was fine under 6.1.x.

jaffacake can you also send your logs?

I have a similar issue, which seems to be very similar to that described. After months of faultless service with a recent firmware upgrade the NAS RN104 goes offline, the power button becomes unresponsive and only works after a power down. Have re-installed the OS. Happens almost every day and even 6.2.2 doesn't fix it.

Hope you can help.

mdgm wrote:
jaffacake can you also send your logs?

Sent

jaffacake your data volume is close to 95% full. I see you have a lot of snapshots. When the volume gets to 95% full we delete the oldest snapshots to free up some space. With a very large number of snapshots this can take quite some time and a lot of IO as each successive snapshot needs to be updated in turn. I would suggest that you backup your data if you haven't already then delete some snapshots.

af8 can you also send your logs?

mdgm wrote:

jaffacake your data volume is close to 95% full. I see you have a lot of snapshots. When the volume gets to 95% full we delete the oldest snapshots to free up some space. With a very large number of snapshots this can take quite some time and a lot of IO as each successive snapshot needs to be updated in turn. I would suggest that you backup your data if you haven't already then delete some snapshots.

Define "a lot of snapshots"? Taking hourly snapshots is always going to add up to a lot of them...24 each day per share. As I understand it, btrfs is very efficient with snapshots so this shouldn't be a major issue.

I'm aware of the 95% issue, we've been here before. In fact, I upgraded due to the apparent improvement in the purging of older snapshots when hitting this threshold. That said, I never hit 95% - I'm generally hovering around the 90% mark right now.

There is no real point in having snapshots if the NAS locks up when I enable them. Likewise, there is little point in having 5 x 3TB drives if the whole thing freezes whenever I use more than 9TB of storage. I defrag, I balance etc. I expect it to work, this isn't a cheap piece of kit.

It was reliable until these latest releases, even though I had lots of data, snapshots etc. Something changed, it wasn't my usage pattern.

jaffacake wrote:
Define "a lot of snapshots"? Taking hourly snapshots is always going to add up to a lot of them...24 each day per share. As I understand it, btrfs is very efficient with snapshots so this shouldn't be a major issue.

The impact of snapshots on file system performance is complex, so I don't think it lends itself to a crisp definition.

Also, "efficient" has more than one dimension. BTRFS snapshots are very space efficient. Make a new snapshot is computationally efficient.

Maintaining snapshots as files are changed is another matter. It can take some time, and it results in file fragmentation. There is a web of reflinks between each snapshot folder (+ the main folder) and the file data on the disk, and every time you change a file those reflinks need to be adjusted (and over time the number of reflinks grows). Maintaining snapshots is space efficient, but performance will drop. How much it drops depends on the usage pattern.

One scenario - you have a media library. You regularly add files, but don't change them, and deletions happen rarely. In this case, snapshots gradually pick up the deleted files. There isn't any fragmentation, because the files aren't being modified. Read and write performance will be good, and shouldn't change much as the number of snapshots grows.

Another scenario - you have a torrent folder. New torrents are being added all the time, and more importantly new torrent pieces are being received every second. You make a snapshot on the hour, right after a torrent starts. A second later, 3 torrent pieces have arrived. The file system has to maintain the original unwritten disk blocks in the snapshot, and link in new disk blocks in the main share for the newly arrived pieces. That takes some work, and the file ends up fragmented. This process continues, and you end up with a completely fragmented file by the time the torrent completes. Read performance and write performance will suffer. Total space usage for each file will end up at least 2x its size, and likely will be more than that if multiple snapshots are taken while the torrent is running (It also depends on the torrent piece size, and how the pieces are aligned with the file system data blocks. Each data block is written at least once after the file is created and initial space allocated, but some data blocks are written more than once because they hold multiple pieces that arrive at different times).

The net here is that snapshots aren't a one-size-fits-all kind of feature. You do need to set up the share options with the expected usage patterns in mind. If you don't do that, then throughput will suffer.

Thanks, I'm aware of that information and have adjusted both snapshot and bitrot/CoW settings accordingly.

That said, a NAS shouldn't just randomly go to sleep. If it was busy, hammering the disks I would understand...it's not. There's no activity, the web browser won't respond and it won't respond to ping. It's the equivalent of a Windows blue screen and it didn't do it before 6.2.x.

Maybe the addition of bitrot protection caused this and it would go away if I disabled? Reality is that my bitrot use is minimal and if it can't handle this basic use then what's the point in implementing the function. This is a device with 6 bays that will allegedly each support 6TB drives. I'm using half that but it appears unable to handle it.

When working, it works very well - I often max out the gigabit link on reads and over 100MB/s on writes. It's great. It just appears to hang and I don't believe it's load related.