January 31, 201412 yr Hey all, Having installed unRAID (5.0.5 Plus) on my new HP Microserver (54L, AMD Turion II, 2GB ECC RAM, everything stock), I think I'm experiencing sub-par performance and would appreciate your feedback. I'm running unRAID off a Lexar Jump Drive S23 USB3.0 stick and the only "aftermarket" addon I have is Unmenu. At first, the array consisted of a WD Green 2TB and a Seagate LP 1.5TB drive (see Disk_setup.JPG for exact model numbers). I did not immediately add a parity drive, as I wanted to migrate my data to the array first (and said data was temporarily stored on a 3TB WD Red which was my soon-to-be-parity drive). As I was transferring data from my PC to the array, I noticed that the write speeds were pretty slow @ around 30MB/s (fluctuating 25-35MB/s). I thought perhaps this was normal (network bottleneck), but then, monitoring memory usage, I noticed that - while things looked pretty normal memory-wise on boot up (with unRAID occupying ~250 megs of RAM) - during transfers there was barely 50MB of free RAM (see HPM_Free and HPM_Free_transfer JPEGs). Again, I thought that this could be normal - perhaps some OS caching. But now, having transferred all files and having added the 3TB Red drive as parity, I run "Tunables Tester" script to fine-tune those tunable values. I was expecting to get a bit of performance boost prior to building the parity (which initially said it'd need 22 hours). To my surprise, the values seem to make no difference, with parity building being capped at 33.6MB/s (see Tunables.JPG)! This does not seem to match the behavior other setups seem to exhibit when tweaking those values (based on what I see from the above-linked thread). All of the above have me a bit worried about my setup and, of course, I would like to troubleshoot any issues before proceeding with the final setup (plugins etc.) Any comments would be welcome as I'm new to unRAID and don't know what to make out of this situation. Thanks in advance!
January 31, 201412 yr I'm no syslog expert, but a few things: Are the SATA ports set in the BIOS to AHCI mode (not IDE). Your flash drive appears to be corrupted. Shut down the server, pull the flash drive, stick it in a PC or Mac and do a check disk. Post SMART reports for the drives.
January 31, 201412 yr Author Thanks for the reply! SATA ports are set to AHCI, yes. I can also confirm that from the most recent syslog (attached). I removed the flash drive, plugged it on my PC, indeed got a "need to run checkdisk"-type of message, run a full scan but found no errors. Still, replugging the flash drive doesn't bring up the "check disk" alert anymore. Plugged the drive back, loaded everything and got the attached SMART reports from the drives. I should mention that the drives passed the pre-clearing process prior to adding them to the array. Hope that helps! syslog-2014-02-01.txt 3TB_WDRed_SMART.txt 2TB_WDGreen_SMART.txt 1.5TB_SeagateLP_SMART.txt
January 31, 201412 yr The flash is not reporting as corrupted in the current syslog, so that's fixed. The 2TB and 3TB drives look fine. The 1.5TB drive is showing a number of Reported Uncorrectable errors. I'd suggest running a long SMART test on it, but be aware this will take a long time.
February 1, 201412 yr Author Yes, I had a few issues with the Seagate drive when it was running on my Windows box (sometimes I could hear it trying to spin up unsuccessfully) but it passed the pre-clearing process with just a few SMART-related warnings and seemed to be working fine. Anyhow, I'll run the full SMART test and get back to you!
February 1, 201412 yr Author I have attached the results of the long SMART test. It seems to pass, even though the same known warnings are reported. 1.5TB_SeagateLP_SMARTb.txt
February 2, 201412 yr The SMART report show the test was aborted by user with 90% of the test remaining. It did not run a full extended test. I'm pretty sure if unRaid spins the drive down it will abort the test. Disable spin down on that drive and start the test again.
February 2, 201412 yr Author Oh, yes, thanks for spotting that. I thought I had set the spin off delay long enough, but apparently that was not the case! Set it to never and I'm currently re-running the test. The only "suspicious" looking entries on my syslog (reported as "Minor Issues" by unmenu) are the following: Feb 1 00:33:06 HP_Microserver kernel: xor: measuring software checksum speed (Minor Issues) ... Feb 1 00:33:06 HP_Microserver kernel: acpi PNP0A08:00: ignoring host bridge window [mem 0x000d0000-0x000dffff] (conflicts with Adapter ROM [mem 0x000cf000-0x000d19ff]) (Minor Issues) The above show up on startup. When unRAID is up and running, I sometimes get this: Feb 1 00:39:57 HP_Microserver emhttp: shcmd (27): :>/etc/samba/smb-shares.conf (Other emhttp) Feb 1 00:39:57 HP_Microserver avahi-daemon[1113]: Files changed, reloading. Feb 1 00:39:57 HP_Microserver emhttp: Restart SMB... (Other emhttp) Feb 1 00:39:57 HP_Microserver emhttp: shcmd (28): killall -HUP smbd (Minor Issues) Feb 1 00:39:57 HP_Microserver emhttp: shcmd (29): cp /etc/avahi/services/smb.service- /etc/avahi/services/smb.service (Other emhttp) (i.e. it kills the smdb service daemon for some reason) Options I'm considering (once the SMART test is done): - Update BIOS on Microserver - Upgrade firmware on the Seagate drive - Upgrade to unRAID 6.0 x64 beta Edit for a comic relief: Apparently Seagate insist that the clicking noise users sometimes notice on Barracuda LP drives is "normal operational noise".
February 2, 201412 yr Author Done. It previously reported: Self-test execution status: ( 25) The self-test routine was aborted by the host. And now I see: Self-test execution status: ( 0) The previous self-test routine completed without error or no self-test has ever been run. So it completed successfully, I guess. Same results, more or less. SMART overall-health self-assessment test result: PASSED I could try installing the new Seagate firmware, though I'm not sure the performance cap I see can be attributed to this drive. Btw, I just initiated a parity check and it runs at ~105MB/s. As soon as it's done, I'll rerun the tunables script to check if the cap is still there. 1.5TB_SeagateLP_SMARTc.txt
February 2, 201412 yr Wait, maybe I misread but was under the impression that you were getting 30-35MB/s writes WITHOUT a parity drive assigned. That would indeed be slow. But if you do have a parity drive assigned, that's about ballpark for write speed. You should be able to read from the array at 100+MB/s regardless. Did I misread?
February 2, 201412 yr Author Wait, maybe I misread but was under the impression that you were getting 30-35MB/s writes WITHOUT a parity drive assigned. Nope, you got it right: the initial post refers to the write performance *without* a parity drive. I tried to ignore the sub-par numbers and, after all the copying was done, went on to add the parity drive. But when I run the tunables script, its results gave me an even clearer indication that something is not right with my setup. Still, I let it build the parity (even with the 35MB/s cap), just for the sake of it.
February 3, 201412 yr Author Options I'm considering (once the SMART test is done): - Update BIOS on Microserver - Upgrade firmware on the Seagate drive - Upgrade to unRAID 6.0 x64 beta Installed the latest HP BIOS on the Microserver (a modded version, to be exact). Not sure if it's related to the BIOS update but I re-run the tunables script and, while a cap is still evident, performance is way better now (~ x3 times the initial one - see attached).
February 5, 201412 yr Author No worries! It's not like I actually resolved anything. ;D Not true! I hadn't spotted the issue with the flash drive. Plus, it's always nice to know you're not just talking to yourself! I think I'll proceed with installing plugins and stuff for now, and resume troubleshooting when I have nothing better to do!
February 18, 201412 yr I am glad that everything is much better now! So now what are your write and read speeds from the NAS (I understand that you have installed 2 disks + a parity disk)? Sent from my iPad using Tapatalk
February 18, 201412 yr Author Two drives + parity + an SSD cache drive, to be exact! Also got me a new WD Red that I'm pre-clearing as we speak. Read speeds are ~70-80MB/s. Write speeds are ~20MB/s when it comes to non-cached shares and ~50+MB/s on cached ones.
February 19, 201412 yr Hm... to tell you the truth I expected some thing more for read & uncached write. For example ~100MB/sec for the read and around 40+ for the write. Since you have a version after 5.0.4, have you tried the optimization mentioned at http://lime-technology.com/forum/index.php?topic=5754.msg289252#msg289252 http://lime-technology.com/forum/index.php?topic=31042.msg279810#msg279810 (enabling some kind of write option) to increase your write speeds. I am asking all these questions because I will have the same setup (without the SSD) in a few days and I am bit disappointed by the write speeds that you mention (30MB/sec)
February 20, 201412 yr Installed the latest HP BIOS on the Microserver (a modded version, to be exact). Can I ask which modded BIOS you loaded? I loaded the version I found from TheBay which comes from AVF... and have disabled IDE mode on the ODD SATA and ESATA ports... the reason I ask is that my write speed appears to be 25MB/s - 30MB/s despite the update. I suspect this is the one you have too as its the only one I have seen so far. At the moment I have some more serious issues which I need to resolve as the unit seems very unreliable at the moment and keeps becoming un-contactable over the LAN.... but once resolved I will need to sort speeds out too! Thanks in advance.
February 21, 201412 yr Author Hm... to tell you the truth I expected some thing more for read & uncached write. For example ~100MB/sec for the read and around 40+ for the write. Since you have a version after 5.0.4, have you tried the optimization mentioned at http://lime-technology.com/forum/index.php?topic=5754.msg289252#msg289252 http://lime-technology.com/forum/index.php?topic=31042.msg279810#msg279810 (enabling some kind of write option) to increase your write speeds. I am asking all these questions because I will have the same setup (without the SSD) in a few days and I am bit disappointed by the write speeds that you mention (30MB/sec) Nope, the only optimization I tried was the tunable settings, but that didn't do much good, as you can see from my earlier posts. Regarding reads, I suspect the bottleneck is in my network, as the drives are read at >100MB/s during other operations. In real use, though, the difference between 75MB/s and 100MB/s is not really noticeable. Writes could be a bit faster, but I don't write to the array that often (transfers are handled via mover transparently), so I'm not that fussed about it either. Things are a bit crazy at work right now, but when I get some free time I may investigate this further for the fun of it. Can I ask which modded BIOS you loaded? Haven't been to TheBay for a while, but I think mine was the newer version (based on the one HP released in early 2013). You can find it here At the moment I have some more serious issues which I need to resolve as the unit seems very unreliable at the moment and keeps becoming un-contactable over the LAN.... but once resolved I will need to sort speeds out too! That's strange. Perhaps something is messed up with the USB stick/unRAID environment? Check the syslog for anything suspicious.
Archived
This topic is now archived and is closed to further replies.