February 27, 201412 yr Hi, I'm hoping to get some insight performance of my new unraid system. I'm a full blown noob. Here's the setup: MB: Supermicro X9SCM-iiF 1230V2 8GB RAM AOC-SAS2LP-MV8 11 drives: On board Sata 3 - Parity (4TB Hitachi 5400K) - Cache (samsung 128GB SSD 830) On board Sata 2: 4TB Hitachi 7200K (in process of preclearing) to replace parity 4TB Hitachi 7200K (in process of preclearing) 3TB Hitachi 5400K - Disk 1 3TB Hitachi 5400K - Disk 2 AOC-SAS2LP-MV8: 3TB Hitachi 5400K - Disk 3 3TB Hitachi 5400K - Disk 4 3TB Hitachi 5400K - Disk 5 2TB WD Green (awaiting assigned) 2TB WD Green (in process of preclearing) Switch is a Netgear prosafe GBe, with Cat6 I'm running a preclear (it is in the post-read phase) on the 3 drives mentioned above. I can transfer to the cache drive @ 93MB/s. I expect a bit better but that's good enough. When I transfer directly to disk 2 i get around 5MB/s. When I transfer to disk 3 i'm getting around 8 with 4 or 5 spikes to 28MB/s. I'm wondering why the speeds are so slow even with the preclear running. Is this typical? It doesn't seem like the controllers should have more room. Thoughts appreciated. Thanks .chris.
February 28, 201412 yr Chris, Going in blindly here as there is no log to go by (i.e. supply one please). I suspect you already disabled INT13 and/or changed out the firmware/BIOS of the MVP card(s)? Check here to verify you did those steps: http://lime-technology.com/forum/index.php?topic=12404.0 Also, is Watchdog enabled on that board? Consider jumper disabling it if so. What BIOS is on the mobo? Revision 2.0a+ seems to resolve a few issues. Everything above and then some is a crap shoot without a log (which no offense, is requested via the sticky in this subforum whenever asking for help for a good reason, it will tell the tale).
February 28, 201412 yr Author Hi grandpix, thanks for the reply. Sorry, here is the syslog. The WatchDog is disabled and I just disabled int13. The bios is at 2.0a. After the preclear finished, I tested transferring a 1GB+ file to each drive. They all basically acted the same, except cache drive. Cache drive was great consistently 100MB/s. All the others would follow the same pattern. Go really fast for the first few seconds (40-60MB/s) and then slow to 2-4MB/s for a while then, ramp back up to 20-30MB/s, then quickly drop back to down 2-4 range, etc. Thoughts? From an untrained eye, I'm guessing the parity drive has some issue although the smart report doesn't show any. I may try switching the ports between the Cache and Parity drive to rule out port specific issue. Thanks for your help! unraidlog.zip
February 28, 201412 yr What is the final average transfer rate? Use Teracopy for testing; it reports the total average rate. Disable all add-ons until the performance issue is resolved.
March 1, 201412 yr Author I decided to swap out my parity drive with my new precleared hitachi 4TB 7200K drive. I started the parity sync and am getting 12-16MB/s as reported in the Dynamix Main page. But if I close the browser, wait a minute or two, then go back to tower/Main... I usually see a much higher number between 80-100MB/s and then watch it fall down to 3-8MB/s I only have dynamix installed with half of their plugins. I'm currently running memtest, once that runs for a bit, I'll disable the plugin and go from there.
March 1, 201412 yr Author memtest ran with no errors. I disabled all the plugins and see a consistent 113-120MB/s parity-sync. that seems right. I'm wondering if Dynamix was reporting it wrong or was causing problems. + 1 dgaschk on using teracopy Now that I've disabled any plugins, i'm seeing speeds I can live with 41MB/s during a parity sync. Hopefully it was something with the plugins... I'll do some more tests tomorrow after the sync finishes.
March 1, 201412 yr Dynamix was reporting it wrong or was causing problems. This. I think there is a setting in Dynamix that disables the constant update that causes the issue. Downside is you have to manually refresh the page to see changes. Or something like that, I don't use Dynamix.
March 1, 201412 yr I have seen slow parity with dynamix. 20MB when I have dynamix on, remove it and I get around 100MB. Dynamix was reporting it wrong or was causing problems. This. I think there is a setting in Dynamix that disables the constant update that causes the issue. Downside is you have to manually refresh the page to see changes. Or something like that, I don't use Dynamix.
March 1, 201412 yr The default setting of Dynamix is real-time page updates, which on certain systems slows down the parity checking (this is due to the way the emhttp implementation of Limetech is done). Changing the page refresh setting to either "passive" or "disabled" restores the parity speed. This is also explained in the Dynamix webGui topic.
Archived
This topic is now archived and is closed to further replies.