Miles- Posted February 6, 2018 Posted February 6, 2018 Since a few weeks the cache pool (JBOD) from my Unraid Server is filling up every day at midnight. After it is completely full it returns to normal and nothing is wrong. I have no idea what is causing the problem. Diagnostics are fetched and can be provided. Thanks in advance. Miles
tdallen Posted February 7, 2018 Posted February 7, 2018 Odd. What is your mover schedule? And what Dockers and plugins do you have running? Do you have any machines elsewhere on your network that might be accessing your server (I.e. a backup) at that time?
Miles- Posted February 7, 2018 Author Posted February 7, 2018 Ifter some testing it has to have something to do with mover. When I started it by hand, I got the same problems. So after looking into the log i found this: I set the prferences to use the chache disks for the system share, but the libvirt folder on the cache disk is empty. In disk 2 there is the libvirt.img. Shouldn't the file be on the cachedrive? Should I move it there?
tdallen Posted February 9, 2018 Posted February 9, 2018 You might want to post your diagnostics. I'm wondering if you have any shares where the data is on the array, and the Cache setting is "Prefer", so mover is trying to move stuff back to the cache drive and running out of space.
trurl Posted February 9, 2018 Posted February 9, 2018 As tdallen said. You almost certainly are confusing cache-yes and cache-prefer settings. See here: https://lime-technology.com/forums/topic/46802-faq-for-unraid-v6/?page=2#comment-537383
Miles- Posted February 11, 2018 Author Posted February 11, 2018 @tdallen , @trurl: You were both right. I changed the Share settings from "YES" to "Prefere". The problem still occured. I manualy copied the file that was causing the error. The error dissapeared. Thank you for your help. SOLVED
trurl Posted February 11, 2018 Posted February 11, 2018 55 minutes ago, Miles- said: @tdallen , @trurl: You were both right. I changed the Share settings from "YES" to "Prefere". The problem still occured. I manualy copied the file that was causing the error. The error dissapeared. Thank you for your help. SOLVED Well, the system share should be cache-prefer. But you don't say which share you set to cache-prefer. And the subject of this thread is cache-disk filling up every night. Any share you have set to cache-prefer is going to be moved to cache when mover runs. Any share you want to be moved from cache should be set to cache-yes. The mistake people make that causes there cache to fill up is setting a share to cache-prefer when it should be set to cache-yes. So, are you sure you have things solved now? Have you run mover to see that everything goes where it should?
Miles- Posted February 12, 2018 Author Posted February 12, 2018 On 11.2.2018 at 1:34 PM, trurl said: Well, the system share should be cache-prefer. But you don't say which share you set to cache-prefer. I set the system share. On 11.2.2018 at 1:34 PM, trurl said: So, are you sure you have things solved now? Have you run mover to see that everything goes where it should? You are right. The Cache disk error disappeared, because Mover is not working at all. This is the error log I'm getting: But on the cache disk are 360GB free space.
trurl Posted February 12, 2018 Posted February 12, 2018 Do you have any other shares that are set to cache-prefer? If you have any shares you want to be written to cache and moved to the array then those shares should be set to cache-yes. So, most shares that use cache should be cache-yes. Only shares you want to be moved to cache and to stay on cache should be set to cache-prefer. On 2/8/2018 at 7:49 PM, tdallen said: post your diagnostics
Miles- Posted February 21, 2018 Author Posted February 21, 2018 Sorry for the late reply. Share - Cache - Settings: appdata -> prefer domains -> prefer isos -> yes media -> yes miles -> yes system -> prefer rest is set to NO nasa-diagnostics-20180221-1909.zip
tdallen Posted February 22, 2018 Posted February 22, 2018 Well, you've actually got several more shares using cache - starting with L, W and X. That said, you have two SSDs, 128GB and 256GB. I am guessing you have them configured for BTRFS RAID-0, spanning with no redundancy? The devices look like they should have sufficient space but there's clearly an issue because the cache drive is reporting full and mover is unable to pull the Cache Prefer files back from the array. I'm thinking maybe something is up with BTRFS and hoping @johnnie.black will check in.
JorgeB Posted February 22, 2018 Posted February 22, 2018 Cache should have lots of space, the syslog ends yesterday: Feb 21 18:46:45 NASA shfs: cache disk full Are you still getting cache full errors? If yes reboot, let the server run for a couple of hours and post new diags.
Miles- Posted February 25, 2018 Author Posted February 25, 2018 On 22.2.2018 at 4:31 PM, tdallen said: Well, you've actually got several more shares using cache - starting with L, W and X. Like I wrote, the rest of the shares is set to NO. On 22.2.2018 at 4:31 PM, tdallen said: That said, you have two SSDs, 128GB and 256GB. I am guessing you have them configured for BTRFS RAID-0, spanning with no redundancy? Correct. On 22.2.2018 at 4:37 PM, johnnie.black said: Cache should have lots of space, the syslog ends yesterday: Feb 21 18:46:45 NASA shfs: cache disk full Are you still getting cache full errors? If yes reboot, let the server run for a couple of hours and post new diags. Yes, still full of chache full errors. Rebooted, let it run an bit, diagnosis in the attachments. nasa-diagnostics-20180225-2030.zip
Squid Posted February 25, 2018 Posted February 25, 2018 Your cache floor settings are wrong. You've got it set to be 2000000000 which is 2TB (kilobytes are assumed when setting the value in the absence of a suffix). Change Settings - Global Share Settings - Cache Device - Cache floor to be 2GB which is probably what you want
JonathanM Posted February 25, 2018 Posted February 25, 2018 40 minutes ago, Miles- said: On 2/22/2018 at 10:31 AM, tdallen said: Well, you've actually got several more shares using cache - starting with L, W and X. Like I wrote, the rest of the shares is set to NO. Cache setting to NO only applies to new files being written to the user share. If the file already exists on the cache, it will not be automatically moved to the array. Turn on the help for that page, and read the descriptions for the different cache settings.
Miles- Posted February 28, 2018 Author Posted February 28, 2018 On 25.2.2018 at 9:05 PM, Squid said: Your cache floor settings are wrong. You've got it set to be 2000000000 which is 2TB (kilobytes are assumed when setting the value in the absence of a suffix). Change Settings - Global Share Settings - Cache Device - Cache floor to be 2GB which is probably what you want This solved the real problem. No more errors. Thank you all so much!
Recommended Posts
Archived
This topic is now archived and is closed to further replies.