June 2, 201115 yr Ok I had a 3 drive system with 3 x 1.5TB WD EADS drives. It was working quite well in a HP Proliant Microserver (after being transferred from running as a VM). I was getting good parity check speeds of 60-65MB/sec (checked before changing config). This was using the free unRAID 4.7. Hardware link is below. It's no super computer but perfectly adequate for unRAID and was cheap at just A$180. Syslog attached upto about 10 minutes after a reboot and parity generation. The 2TB drives were fitted to pre-clear for a while and that went along no problem. http://h10010.www1.hp.com/wwpc/us/en/sm/WF05a/15351-15351-4237916-4237918-4237917-4248009.html I wanted more space so I paid for the Plus key and bought 2 x 2TB Seagate Green drives (eventual plan to have 3x1.5TB + 2x2TB for 6.5TB of space). So first thing was to replace the parity drive with one of those 2TB drives. That's where I'm at now. I stopped the array, replaced the parity drive (with the larger 2TB) and restarted it after checking the "are you sure?" box. So now I have (as an intermediate stage) 2TB - parity 2 x 1.5TB - storage Parity generation speed is jumping around but is usually less the 5MB/sec but sometimes briefly jumps up to around 10-15MB/sec. e.g. currently Estimated speed: 2,647 KB/sec Estimated finish: 12278.5 minutes The two 1.5TB data drives seem to be seeking about quite a bit whereas the parity one isn't (can hear and feel them). They are different drives so it's hard to say what the newer parity drive is doing. I did temporarily go back to 3 x 1.5TB and it was even a bit slower than this. There are no hardware type errors in the syslog. The 2TB gives 120MB/sec+ speeds for reading and writing. The 2x1.5TB give about 75-80MB/sec read speed (can't write since they have the current file system on them) This seems excessively slow and current targets put this about 10 days away with no protection in the mean time Any ideas? I don't know what's changed that could cause this. Thanks Christian syslog.txt
June 2, 201115 yr Have you tried reseating the SATA cables? Maybe one is loose and slowing down data transfer speeds. Did you preclear your new drive? Can you check the SMART reports of the drives to see if there are any issues? Sorry - I'm not expert here, so these suggestions might be useless... just trying to use what I've learned so far
June 2, 201115 yr Author Thanks for the suggestions. I had already fitted the 2TB drives and it was working OK and didn't physically touch the machine since. Also sda through sdd are all in a cage with only one cable. I'd expect lots of errors in the syslog if there was a cable problem or worst case the drives would simply not be detected. I did pre-clear the new 2TB drive although I don't think that is as significant for a parity drive aside from the pre-stressing side of it. The other two 1.5TB drives do have high load cycle counts in SMART (damn WD) but otherwise are working fine. One does have "» load_cycle_count=336976 » current_pending_sector=4 » offline_uncorrectable=1 » multi_zone_error_rate=14" in the unMENU status page. I know this isn't good and will probably have to retire these drives soon but in the mean time I just want my protected array going in the interim.
June 2, 201115 yr Yes - that drive isn't looking to healthy. I'm not sure how wise it is to put a known bad drive into the array. I think if I were you, I'd copy the data from the bad drive to a new one first to make sure it doesn't die with all your data on it. Then go into building parity with known good hardware. Honestly not sure what is causing your slow parity build speeds... unlikely to be the drive I would have thought, but perhaps it's a hint from a "higher power" to replace the drive.
June 2, 201115 yr Author I did link to the hardware specs - it's an off the shelf system - just added drives and USB key http://h10010.www1.hp.com/wwpc/us/en/sm/WF05a/15351-15351-4237916-4237918-4237917-4248009.html http://h18004.www1.hp.com/products/quickspecs/13716_na/13716_na.html Processor Dual-Core Processor AMD Athlon II Processor Model NEO N36L (1.30 GHz, 15W, 2MB) Cache Memory 2x 1MB Level 2 cache Chipset AMD RS785E/SB820M Memory Protection ECC Memory Type PC3-10600E unbuffered DDR3 ECC operating at 800MHz 1GB (1x1GB) Network Controller Embedded NC107i PCI Express Gigabit Ethernet Server Adapter 1 PCI-e x16 x16 1h Half height, half length slot 2 PCI-e x1 x1 2h Half height, half length slot Storage Controller Integrated SATA controller with embedded RAID (0, 1) Interfaces Graphics On board VGA USB 2.0 Ports 7 total (2 rear, 4 front panel, 1 internal) Network RJ-45 (Ethernet) 1 (10/100/1000 bits/s) eSATA Gen 2 1 rear Graphics 128MB shared supporting 1920x1200 @ 60Hz Industry Standard Compliance ACPI V2.0 Compliant PCI 2.3 Compliant PXE Support WOL Support IPMI 2.0 compliant USB 2.0 SATA Gen 2 Parity check *before* replacing the drive was looking like this from the automated emails I have setup: Parity CHECK/RESYNC in progress, 47.1% complete, est. finish in 187.2 minutes. Speed: 68953 kb/s. Parity CHECK/RESYNC in progress, 62.4% complete, est. finish in 140.1 minutes. Speed: 65436 kb/s. Parity CHECK/RESYNC in progress, 77.7% complete, est. finish in 87.3 minutes. Speed: 62312 kb/s. After the drive replacement: Parity CHECK/RESYNC in progress, 1.2% complete, est. finish in 6752.1 minutes. Speed: 4757 kb/s. Parity CHECK/RESYNC in progress, 3.1% complete, est. finish in 2147.3 minutes. Speed: 14680 kb/s Parity CHECK/RESYNC in progress, 4.8% complete, est. finish in 2277.5 minutes. Speed: 13602 kb/s. Parity CHECK/RESYNC in progress, 5.8% complete, est. finish in 12273.2 minutes. Speed: 2493 kb/s. Parity CHECK/RESYNC in progress, 6.7% complete, est. finish in 12582.1 minutes. Speed: 2412 kb/s Parity CHECK/RESYNC in progress, 7.9% complete, est. finish in 12286.9 minutes. Speed: 2435 kb/s. Parity CHECK/RESYNC in progress, 8.9% complete, est. finish in 11670.4 minutes. Speed: 2536 kb/s. (these are at hourly intervals)
June 3, 201115 yr Author If it's a SMART type issue then why did it all of a sudden drop in performance? I got it cheap I think because of a reseller error but maybe it was a clearance too. It was quite a good deal for a solid little 5 bay device. There was a massive rush over a few days to get them. One reseller even bought a bunch o stock and then resold them. smartlogs_sda.txt smartlogs_sdb.txt smartlogs_sdc.txt
June 3, 201115 yr It would drop all of a sudden if a drive starts to fail. SMART could show this. Th drives look ok but the Seagates Seek_Error_Rate my be causing the problem. Do you have a spare you can swap in?
June 3, 201115 yr Author The Seagate is absolutely brand new. Never installed anywhere else. I ran the pre clear script on it before using it too. I do have another one installed (currently sde) that I'm running the pre-clear on today. Results attached from the current one. However when I first noticed the very poor parity build speed, I stopped and went back to using the original parity drive (was sdd) and got similar slow speeds (actually a bit slower). So that kinda rules out the Seagate. Is there a significant difference between a parity build and a parity check apart from writing the parity drive vs reading the parity drive? That's about the only thing I can see that has significantly changed. In fact, I still even get quite good hdparm results even *whilst* doing the parity build. root@Mars:~# hdparm -Tt /dev/sd[abc] /dev/sda: Timing cached reads: 1540 MB in 2.00 seconds = 770.26 MB/sec Timing buffered disk reads: 318 MB in 3.01 seconds = 105.55 MB/sec /dev/sdb: Timing cached reads: 1118 MB in 2.00 seconds = 558.89 MB/sec Timing buffered disk reads: 86 MB in 3.22 seconds = 26.74 MB/sec /dev/sdc: Timing cached reads: 998 MB in 2.00 seconds = 498.66 MB/sec Timing buffered disk reads: 96 MB in 3.09 seconds = 31.09 MB/sec preclear_start__5YD2SZRE_2011-06-02.txt preclear_finish__5YD2SZRE_2011-06-02.txt preclear_rpt__5YD2SZRE_2011-06-02.txt
June 5, 201115 yr Author Update: I noticed that the load cycle count had increased significantly. I had just recently turned off the automatic head parking using the wdidle3 tool. It had been set to 8 seconds previously. "Disabling" actually seemed to make it much much worse. One drive went from about 300k cycles to about 600k whilst trying rebuild the parity over a few days (basically since my first post). I could see the count going up just by refreshing the unmenu SMART page. So I cancelled the parity check, rebooted a DOS session to run the wdidle3 program again but this time I set it to 5 minutes rather than disabled. After rebooting, the speed started around 16000kB/sec and now is running at a fairly consistent 21000kB/sec (rather than jumping up and down like it was before). 30% complete after 7.5 hours. Also the load cycle counts are no longer increasing. Still not super fast but is this speed of ~20MB/sec reasonable to see (it will take <1 day to complete instead of 10-20)? I don't remember how long it took originally when I set it up. Thanks for listening.
June 5, 201115 yr It's still pretty slow. Just swapping a drive should not have caused such a change in parity build time. Any errors in the syslog during the parity build? Peter
June 6, 201115 yr Author There was one hard error on one of the drives in the syslog but that shouldn't have caused it to run slow the whole time (syslog from boot attached). The automated status emails I got showed Parity CHECK/RESYNC in progress, 01.7% complete, est. finish in 1508.5 minutes. Speed: 21197 kb/s. Parity CHECK/RESYNC in progress, 05.8% complete, est. finish in 1439.7 minutes. Speed: 21277 kb/s. Parity CHECK/RESYNC in progress, 09.9% complete, est. finish in 1235.1 minutes. Speed: 23734 kb/s. Parity CHECK/RESYNC in progress, 14.0% complete, est. finish in 1200.4 minutes. Speed: 23299 kb/s. Parity CHECK/RESYNC in progress, 18.3% complete, est. finish in 1163.0 minutes. Speed: 22841 kb/s. Parity CHECK/RESYNC in progress, 22.5% complete, est. finish in 1157.9 minutes. Speed: 21764 kb/s. Parity CHECK/RESYNC in progress, 26.7% complete, est. finish in 1069.1 minutes. Speed: 22317 kb/s. Parity CHECK/RESYNC in progress, 30.8% complete, est. finish in 1051.7 minutes. Speed: 21416 kb/s. Parity CHECK/RESYNC in progress, 34.7% complete, est. finish in 920.3 minutes. Speed: 23065 kb/s. Parity CHECK/RESYNC in progress, 38.7% complete, est. finish in 952.9 minutes. Speed: 20917 kb/s. Parity CHECK/RESYNC in progress, 42.5% complete, est. finish in 850.1 minutes. Speed: 21981 kb/s. Parity CHECK/RESYNC in progress, 46.6% complete, est. finish in 740.2 minutes. Speed: 23479 kb/s. Parity CHECK/RESYNC in progress, 50.5% complete, est. finish in 755.2 minutes. Speed: 21300 kb/s. Parity CHECK/RESYNC in progress, 54.4% complete, est. finish in 714.7 minutes. Speed: 20734 kb/s. Parity CHECK/RESYNC in progress, 58.3% complete, est. finish in 629.6 minutes. Speed: 21524 kb/s. Parity CHECK/RESYNC in progress, 62.4% complete, est. finish in 559.9 minutes. Speed: 21856 kb/s. Parity CHECK/RESYNC in progress, 66.1% complete, est. finish in 517.2 minutes. Speed: 21290 kb/s. Parity CHECK/RESYNC in progress, 69.9% complete, est. finish in 484.8 minutes. Speed: 20188 kb/s. Parity CHECK/RESYNC in progress, 73.6% complete, est. finish in 412.9 minutes. Speed: 20733 kb/s. Parity CHECK/RESYNC in progress, 81.7% complete, est. finish in 91.2 minutes. Speed: 65055 kb/s. Parity CHECK/RESYNC in progress, 92.7% complete, est. finish in 40.9 minutes. Speed: 57647 kb/s. These are spaced hourly so speed was consistent to 75% and then it got much faster. I assume that's because currently the data drives are only 1.5TB each and the parity is 2TB so the last 0.5TB of parity isn't really anything (yet) So now the parity generation has completed I just started a parity check. It is running from the start at about 63MB/sec which is what was showing before I made all the changes. At least that shows there probably isn't anything to do with the disk change. So the question is why is the parity *build* about one third the speed of the parity *check*? Thanks syslog.txt
Archived
This topic is now archived and is closed to further replies.