kenmaglio Posted August 14, 2015 Share Posted August 14, 2015 I have (a rather slow) server which is running 13 disks. 1 parity, 11 data, 1 cache ssd. My parity calculation is really slow 15mb/sec. While I thought this was going to be okay for the initial sync, it looks as if when I add something like a single folder, it recalculates all of the parity - not just what's changed? Is this correct behavior? Maybe my parity wasn't completely finished (my webgui was broken - so I could not see for sure before I restarted, but there was no disk activity). How is a recalculation of parity done when a file is dropped onto the disk pool via a user Share? Is it really going to recalculate parity entirely -- or should it just recalc the sections of the disks that changed... Edit: When I setup the array without parity for initial copy, I was getting near 100mb writing to one disk. Link to comment
kenmaglio Posted August 14, 2015 Author Share Posted August 14, 2015 I looked at my syslog and saw this: this mean it wasn't finished and I wasted all that time calculating the parity?? am I correct in that theory? Aug 13 18:50:33 unRaid kernel: md: recovery thread woken up ... Aug 13 18:50:33 unRaid kernel: md: recovery thread syncing parity disk ... It takes like 4 days for my parity to complete at 15mb sec. Link to comment
RobJ Posted August 14, 2015 Share Posted August 14, 2015 It only recalculates for the changes. It's not clear at all what your situation is. The indications are that parity had not been built before. Please provide the Diagnostics zip, see Tools -> Diagnostics. Link to comment
kenmaglio Posted August 14, 2015 Author Share Posted August 14, 2015 Here ya go unraid-diagnostics-20150813-1907.zip Link to comment
RobJ Posted August 14, 2015 Share Posted August 14, 2015 I looked at my syslog and saw this: this mean it wasn't finished and I wasted all that time calculating the parity?? am I correct in that theory? Aug 13 18:50:33 unRaid kernel: md: recovery thread woken up ... Aug 13 18:50:33 unRaid kernel: md: recovery thread syncing parity disk ... It takes like 4 days for my parity to complete at 15mb sec. Aug 13 18:50:33 unRaid kernel: mdcmd (49): check CORRECT Aug 13 18:50:33 unRaid kernel: md: recovery thread woken up ... Aug 13 18:50:33 unRaid kernel: md: recovery thread syncing parity disk ... Your 2 lines are preceded by the 'check CORRECT' line, which would normally indicate that's not a parity build, but a parity check, and since I don't see any indication of an unclean shutdown, it almost looks like a parity check that you may have started? However, elsewhere it says parity is Invalid, so perhaps it IS a parity build. It's logged a little unusual. Perhaps the previous parity build had not completely finished? The reason the parity operation is slow is the 2 port multipliers. In fact, the 8 drives on them took so long showing up, they almost missed roll call! Plugins had been loaded, and the go file run before the port multipliers and their drives showed up, seconds before unRAID checked the inventory of drives. Single drive access to any of those drives should be OK, but once a parity check or build starts, then bandwidth is severely throttled by the port multiplier connections, making it VERY slow. Apart from that, there are no other visible indications why it's slow. Link to comment
kenmaglio Posted August 14, 2015 Author Share Posted August 14, 2015 In fact, the 8 drives on them took so long showing up, they almost missed roll call! Plugins had been loaded, and the go file run before the port multipliers and their drives showed up, seconds before unRAID checked the inventory of drives. Should I add additional sleep wait into the Go script after the drivers are loaded? echo Starting RocketRaid AHCI /boot/scripts/enable_ahci.sh echo Waiting 5 seconds sleep 5 echo Start the Management Utility /usr/local/sbin/emhttp -p 8088 & Should I increase to 15 lets say? Link to comment
RobJ Posted August 14, 2015 Share Posted August 14, 2015 I'd forgotten about that 'sleep 5'! That explains the slightly more than 3 second gap, where nothing at all appears to happen. That 5 might be long enough, but would be safer to add a few more seconds, for insurance. Maybe 'sleep 8'. Link to comment
kenmaglio Posted August 16, 2015 Author Share Posted August 16, 2015 So I have built a new server. Unfortunately not what I wanted to do, but I had to. I went from a crappy and understandably why, 15mb/sec on parity to now 150mb/sec. Full parity of 10 drives, 40 TB, is now only about to take 9 hours. Link to comment
RobJ Posted August 17, 2015 Share Posted August 17, 2015 I went from a crappy and understandably why, 15mb/sec on parity to now 150mb/sec. Wow! Money is tight for some of us too. When you read the forums, it may seem like everyone else has a huge system, and can throw money around at every new toy, but for many of us, probably quieter posters, improvements come with sacrifice. I hope you will feel this one was worth it! Link to comment
kenmaglio Posted August 17, 2015 Author Share Posted August 17, 2015 Rob, Yes I'm very, very happy, complete parity calculation against 10 4TB drives -- took only 9hrs. To be honest I didn't go crazy, but I did not spend that little either, near 1k - for an entirely new system + cabling + 1 new 4tb wd red (had a crap drive in there I wanted gone -- now all 10 are WD reds!). The hardest thing was finding an enclosure to hold it all. Nanoxia Deep Silence 6 (no link - only in store -- $220) mfg website Corsair 750 (for the power connections - 8 sata, 4 molex) link Gigabyte link A10 link 32GB Crucial link Link to comment
Recommended Posts
Archived
This topic is now archived and is closed to further replies.