November 17, 20169 yr My server has been performing like a 1st gen pentium the last week or so - disk transfers take forever or fail, plex converts files at 0.1x speed so one file per day, cant watch anything as buffers, parity check does 0.3% after 8 hours and so on can someone help me get my super-fast machine back please highlander-diagnostics-20161117-1712.zip
November 17, 20169 yr Author See here: https://lime-technology.com/forum/index.php?topic=41588.0 I've already tried restarting several times
November 17, 20169 yr shareCacheFloor="25GB" One quick response - your cache floor is set at 25GB, but I believe that value is not in bytes but in kilobytes, so '25GB' represents a floor of 25TB, a bit excessive! That would make your Cache drive look full even if empty. Check the Help on that value. It would be a good idea for the code to include error checking here, perhaps disallow numbers greater than half the drive size, at least force it to be less than 80% of drive size.
November 17, 20169 yr It would be a good idea for the code to include error checking here, perhaps disallow numbers greater than half the drive size, at least force it to be less than 80% of drive size. If you want to get really fancy, a slider control from 0% to 80%(or whatever) would be dandy. It would totally eliminate unit errors, and get to the heart of what most people intend to do.
November 17, 20169 yr Author shareCacheFloor="25GB" One quick response - your cache floor is set at 25GB, but I believe that value is not in bytes but in kilobytes, so '25GB' represents a floor of 25TB, a bit excessive! That would make your Cache drive look full even if empty. Check the Help on that value. It would be a good idea for the code to include error checking here, perhaps disallow numbers greater than half the drive size, at least force it to be less than 80% of drive size. are you sure??! This represents a "floor" of the amount of free space remaining on the cache disk. If the free space becomes less than this value, then new files written to user shares with cache enabled will go to the array and not the cache disk.Enter a numeric value with one of these suffixes: KB = 1,000 MB = 1,000,000 GB = 1,000,000,000 TB = 1,000,000,000,000 If no suffix, a count of 1024-byte blocks is assumed. Examples: 2GB => 2,000,000,000 bytes 2000000 => 2,048,000,000 bytes
November 17, 20169 yr are you sure??! Does changing it fix your shfs/user: share cache full problem? That Help text is relevant to what you type into the WebGUI, not necessarily what is stored in the share.cfg file.
November 17, 20169 yr Author Mover constantly running has been one of problems. I've stopped all dockers to see if mover finishes and then I'll stop the array to change the setting
November 17, 20169 yr No, you're right. shareCacheFloor="25GB" is valid. Nov 15 23:46:47 Highlander emhttp: shcmd (53): /usr/local/sbin/shfs /mnt/user -disks 63 25000000000 -o noatime,big_writes,allow_other -o remember=0 |& logger As a comparison, I've just changed my Cache Floor shareCacheFloor="20GB" and I have Nov 17 19:36:23 Mandaue emhttp: shcmd (213238): /usr/local/sbin/shfs /mnt/user -disks 63 20000000000 -o noatime,big_writes,allow_other,use_ino -o remember=330 |& logger It might be worth checking for file system corruption on your cache disk.
November 17, 20169 yr Author No, you're right. shareCacheFloor="25GB" is valid. Nov 15 23:46:47 Highlander emhttp: shcmd (53): /usr/local/sbin/shfs /mnt/user -disks 63 25000000000 -o noatime,big_writes,allow_other -o remember=0 |& logger It might be worth checking for file system corruption on your cache disk. Gonna nuke the cache and rebuild
November 17, 20169 yr are you sure??! Nope! That came from memory, what I thought was correct, but I can't find mention of it anywhere, in docs or manuals. Somewhere, there is a configurable number stored in KB, but it may not be this one. If it's truly in bytes, then the current default of "2000000" (less than 2MB) seems far too small. I had always thought that that represented almost 2GB. I do like the idea of a slider!
November 17, 20169 yr From the Help, "2000000" represents 2048000000 bytes, while "2GB" represents 2000000000 bytes. I'm interested to see that while we both have fuse_remember="330" and I have -o remember=330 , you have -o remember=0 . I wonder if that's relevant?
November 18, 20169 yr Author OK removing and resetting the cache seems to have done the trick - I think somehow I'd managed to get the mover in a tiss, by moving a lot of files it was trying to move at the same time so the machine came to almost a halt.
November 18, 20169 yr Author OK removing and resetting the cache seems to have done the trick - I think somehow I'd managed to get the mover in a tiss, by moving a lot of files it was trying to move at the same time so the machine came to almost a halt. hmm not quite. It still seems to not be respecting my 25GB cache setting and going below. highlander-diagnostics-20161118-0908.zip
November 19, 20169 yr Author Grrr this is becoming unbearable -I can barely even type now my machine has got so bad. highlander-diagnostics-20161119-2141.zip
November 19, 20169 yr Have you tried removing the cache from the equation by setting your user shares to "Use cache disk: No" and stopping any dockers and VMs? You need to break the problem into smaller pieces and I would start by eliminating the cache.
November 19, 20169 yr Author I kind of did that - stopped everything, rebuilt cache and all was good for around 24 hours
November 19, 20169 yr It's not really the same thing. I think you need to get it working as a plain NAS again with all the extras stripped out. Once it's stable you can start adding them back. There's a lot depending on that cache and I think you'll struggle to get to the bottom of the issue unless you break it into manageable pieces. Well, that's what I'd do, anyway.
November 21, 20169 yr Author ok I think I've found the problem: - In cache settings, I had min free space for cache set to 25GB - for some shares like media, I'd followed the advice and set to bigger than the largest file I'd transfer at 30GB What I think was happening is that my cache at say 31GB free would allow Unraid to start transferring a 10GB file that met the 30GB criteria but failed the 25GB criteria, and would get in a tizz and stuck and trying to run the mover to fix just made matters worse. I've now set my share settings to be the biggest file I'd transfer on TOP of the 25GB i.e. 50GB now and no 25GB. It means I'm potentially not using 20% of my 250GB cache so if it continues to work, I'll set the cache setting to 5GB and the share to 30GB to reduce wastage.
November 21, 20169 yr Community Expert I do not think the problem is quite what you describe? The Min Free Space limits only take effect at the start of a file transfer when the new file is created, and take no account of the size of the file. Once unRAiD has selected a disk it does not change its mind if the file proves too large to fit in the available space, but reports an error instead. That is why you must make it large enough to allow for the largest file to be transferred if you do not want the transfer to error out. If you are running parallel transfers then you have to allow for the maximum space they all require added together. When the transfers complete the free space can be well below the value specified depending on how far the sizes of the files in transit at the time took it below that value. However it must never reach zero or the transfers error out.
November 23, 20169 yr Author I do not think the problem is quite what you describe? The Min Free Space limits only take effect at the start of a file transfer when the new file is created, and take no account of the size of the file. Once unRAiD has selected a disk it does not change its mind if the file proves too large to fit in the available space, but reports an error instead. That is why you must make it large enough to allow for the largest file to be transferred if you do not want the transfer to error out. If you are running parallel transfers then you have to allow for the maximum space they all require added together. When the transfers complete the free space can be well below the value specified depending on how far the sizes of the files in transit at the time took it below that value. However it must never reach zero or the transfers error out. Maybe, but I've had no problems since making this change so the problem was with the cache either via the mover, recycle bin playing up, or something else
Archived
This topic is now archived and is closed to further replies.