April 10, 201313 yr Hi, I started a pritiy check to my array. The Parity drive is WD 2TB Green, with 20k hours (2+ years) The are a lot of correction since threre was a drive missing for 2 weeks and copies were made so the missing drive so I had to rebuild parity. I started and went to sleep... woke up and saw only 15GB was checked so far... Strange.. the speed is 300KB/s I stopped, rebooted and ran again.. Speed was 100-110 MB/s now... up untill the parity check started correcting... (15GB point.) Did this about 3 times over again and same result... fast until write is needed. I ran a short smart test, all good. Its running the long now.. It might be a bad drive.. Stupied question: I noticed that whenever a correction is made the log is written. I guess the log is on the unRaid USB. I ran a command for mdcmd couple of months ago chaning the log report level. Can this effect the speed? Apr 11 01:22:53 Storage kernel: md: correcting parity, sector=31108744 (unRAID engine) Apr 11 01:22:53 Storage kernel: md: correcting parity, sector=31108752 (unRAID engine) Apr 11 01:22:53 Storage kernel: md: correcting parity, sector=31108760 (unRAID engine) Apr 11 01:22:53 Storage kernel: md: correcting parity, sector=31108768 (unRAID engine) Apr 11 01:22:53 Storage kernel: md: correcting parity, sector=31108776 (unRAID engine) Apr 11 01:22:53 Storage kernel: md: correcting parity, sector=31108784 (unRAID engine) Apr 11 01:22:53 Storage kernel: md: correcting parity, sector=31108792 (unRAID engine) Apr 11 01:22:53 Storage kernel: md: correcting parity, sector=31108800 (unRAID engine) Apr 11 01:22:53 Storage kernel: md: correcting parity, sector=31108808 (unRAID engine) Apr 11 01:22:53 Storage kernel: md: correcting parity, sector=31108816 (unRAID engine) Apr 11 01:22:53 Storage kernel: md: correcting parity, sector=31108824 (unRAID engine) Apr 11 01:22:53 Storage kernel: md: correcting parity, sector=31108832 (unRAID engine) Apr 11 01:22:53 Storage kernel: md: correcting parity, sector=31108840 (unRAID engine)
April 12, 201313 yr Author Hi again... The long SMART test was good, no errors. I saw that mdcmd stopped logging after several corrections, so it not the issue . Apr 12 07:08:29 Storage kernel: md: correcting parity, stopped logging (unRAID engine) I tried coping to each disk directly (/mnt/disk1 , /mnt/disk2 etc...) this was ok, I got 30MBs... This is also the copy for the parity... so I dont get it... How a direct copy goes fine but a parity check for WRITES only gives me 300KBs. The estimation is horriable, 28 days for a parity check... Total size: 2 TB Current position: 85.73 GB (4%) Estimated speed: 335.62 KB/sec Estimated finish: 94482 minutes Sync errors corrected: 1889070 Can someone help?
April 13, 201313 yr Author Realy desperate here... Here is some more detail: There is a constant read, write , write cycle... I can see it through unMenu`s Disk Performance gui: Cycle1: READ Device Read Write Total sde: 0.00 KB/s 166.01 KB/s 166.01 KB/s sdg: 166.01 KB/s 0.00 KB/s 166.01 KB/s sdd: 166.01 KB/s 0.00 KB/s 166.01 KB/s sdi: 166.01 KB/s 0.00 KB/s 166.01 KB/s sdh: 166.01 KB/s 0.00 KB/s 166.01 KB/s sdf: 166.01 KB/s 0.00 KB/s 166.01 KB/s total: 830.04 KB/s 166.01 KB/s 996.05 KB/s Cycle2: WRITE Device Read Write Total sde: 0.00 KB/s 124.76 KB/s 124.76 KB/s sdg: 0.00 KB/s 0.00 KB/s 0.00 KB/s sdd: 0.00 KB/s 0.00 KB/s 0.00 KB/s sdi: 0.00 KB/s 0.00 KB/s 0.00 KB/s sdh: 0.00 KB/s 0.00 KB/s 0.00 KB/s sdf: 0.00 KB/s 0.00 KB/s 0.00 KB/s total: 0.00 KB/s 124.76 KB/s 124.76 KB/s Cycle3: WRITE Device Read Write Total sde: 0.00 KB/s 166.34 KB/s 166.34 KB/s sdg: 0.00 KB/s 0.00 KB/s 0.00 KB/s sdd: 0.00 KB/s 0.00 KB/s 0.00 KB/s sdi: 0.00 KB/s 0.00 KB/s 0.00 KB/s sdh: 0.00 KB/s 0.00 KB/s 0.00 KB/s sdf: 0.00 KB/s 0.00 KB/s 0.00 KB/s total: 0.00 KB/s 166.34 KB/s 166.34 KB/s The cycle is actually the interval from the unMenu command, I think 1 Second. Also, this is a VM, using an M1015 (IT-MODE) passed through card. The processor is XEON 1240-V2 Average CPU usage is 233 MHZ Average Mem is about 120MB (of Dedicated 2GB) Any hope?
April 13, 201313 yr Hi again... The long SMART test was good, no errors. I saw that mdcmd stopped logging after several corrections, so it not the issue . By design, it only logs the first 20 errors it finds. The remaining, if any, do not get printed to the log. Attach to the next post you make a zipped copy of your syslog. Nobody can help you without the details for analysis. All we can do without it is agree with you, it is slow and probably something is either broken or not configured properly.
April 13, 201313 yr Author Hi, Thank you for the reply. Here is the system log. syslog-2013-04-13.zip
April 18, 201313 yr bump... Any clues, directions? I`m stil unproteceted. I did not see anything specific in the syslog other than it logged the first 20 of parity disk corrections (all it ever logs) Just let the parity check continue...
April 18, 201313 yr Author I did, it takes ages and dosent speed up. 30 days parity check is something I might make once but I cant relay on a system that acts like this. I need to isolate the problem. If nothing helps I will have to move it to a dedicated workstation and check...
April 19, 201313 yr I did, it takes ages and dosent speed up. 30 days parity check is something I might make once but I cant relay on a system that acts like this. I need to isolate the problem. If nothing helps I will have to move it to a dedicated workstation and check... I agree, that is WAY too slow. Just have no experience with VP or pass-through, but it sure sounds inefficient.
April 19, 201313 yr Author I`v wasted a whole friday on this... Set a dedicated workstation, replaced my SAS card with another... Ton of work... no solution. With no hope I degraded to unraid 5.0-rc10 (was 5.0-rc12a)... What do you know... it works now. 34.5 MB/s Now, I have tested this on a workstation so virtualization has no part here... Did anyone do a complete Parity check with LOTS of correction? Maybe no one tested it.... so its a bug in 12a I`m pretty sure. I tried everything... only rc10 worked.
April 19, 201313 yr You haven't posted details of your hardware (and your syslog was far from complete), so it's difficult to be sure, but ... it seems to me, that this may be an example of the slow writes which have affected several systems based on the Supermicro X9SCM motherboard. Have you tried restricting available memory to 4GB? Have you tried setting 'sysctl vm.highmem_is_dirtyable=1'?
April 19, 201313 yr Author Thanks for commenting... The syslog posted was the whole code from unMenu. My HW: Supermicro X9SCM-F bios 2.0a, 2x IBM M1015 passthrough (ZFS & unRaid), 1x GPU passthrough (win), 24 GB Mem The other suggestions was not applied. Are these mandatory for rc12a? rc10 works fine.
April 20, 201313 yr The syslog posted was the whole code from unMenu. Sorry, ignore my comment - you have indeed, posted a full syslog. My ageing memory had linked your post with that of another user! My HW: Supermicro X9SCM-F bios 2.0a, 2x IBM M1015 passthrough (ZFS & unRaid), 1x GPU passthrough (win), 24 GB Mem The other suggestions was not applied. Are these mandatory for rc12a? rc10 works fine. I'm guessing that you haven't seen this thread.
April 20, 201313 yr Author My unRaid VM has 4GB of mem allocated (all dedicated due to VMDP req). Im not passing the 4GB bar... I did not experiance slow writes since installed rc12... only slow Parity writes... Note that this happen while in workstation mode... I took out the ESXi usb and replaced it with the unRaid one. This loaded the unRaid machine without issues... Did the Parity check on that.. still slow. Another thing is the version, the post creator referenced rc8a, not 12a so this is not the case here. Maybe it is a local issue to my setup... tough I must say its a common setup here. I guess if this is a global issue people will start asking questions. For now i`ll stick with rc10, it works. Thanks!
Archived
This topic is now archived and is closed to further replies.