LinHES Forums http://forum.linhes.org/ |
|
XFS new HD recordings have large number of extents & choppy http://forum.linhes.org/viewtopic.php?f=23&t=24116 |
Page 1 of 1 |
Author: | thekingofspain [ Mon Dec 22, 2014 7:05 pm ] |
Post subject: | XFS new HD recordings have large number of extents & choppy |
I have been running LinHES for a few years on RAID 1 consisting of two 1.5 TB Sata drives. Over the last 6 months playback has been getting choppier and choppier at times. The seems to to be patterns to the trend:
2. A recording with a large number of extents. (reducing extents fixes issue unless #1 also occurring) When either issue above causes playback issues, myth front end reports waiting for ringbuffer errors. Watching the playback results always points to not enough data getting but into the hopper. iostat also reveals 100% usage, but at other times with 100% usage there are no issues during playback. CPU is %30 or less with one core at near 100% and the other three at under 5%. My 1.5 TB Raid is about 75% filled with a max fragmentation of 5% fragmentation. I have performed some cleaning, was 85% filed, to increase the percentage of freespace and lower the extent fragmentation on the existing recording. I believe I had about 10 recording with 10 extents, it just went down from there. After this cleaned up, three shows recorded today with two recording at the same time. I do not if the show content is or good or not at this point, but the overlapping shows had 24 extents in them and the solo show had 22 extents. Running xfs_fsr reduced each file to 1 extent. Here is some system info: /etc/fstab /dev/md0 /data/storage/raid xfs defaults,inode64,noatime,nodiratime,attr2,nosuid,nodev,allocsize=64m,filestreams 0 0 newly added settings hdparm /dev/sd<b/c/> /dev/sd<b/c>: multcount = 0 (off) IO_support = 1 (32-bit) readonly = 0 (off) readahead = 256 (on) geometry = 182401/255/63, sectors = 2930277168, start = 0 mdadm --detail /dev/md0 /dev/md0: Version : 1.2 Creation Time : Thu Sep 1 01:08:19 2011 Raid Level : raid1 Array Size : 1465137424 (1397.26 GiB 1500.30 GB) Used Dev Size : 1465137424 (1397.26 GiB 1500.30 GB) Raid Devices : 2 Total Devices : 1 Persistence : Superblock is persistent Update Time : Mon Dec 22 16:22:53 2014 State : clean, degraded Active Devices : 1 Working Devices : 1 Failed Devices : 0 Spare Devices : 0 Name : mythtv:0 UUID : e9674e12:a891594b:39404e66:ecf488b6 Events : 45534 Number Major Minor RaidDevice State 0 0 0 0 removed 1 8 16 1 active sync /dev/sdb xfs_info /dev/md0 meta-data=/dev/md0 isize=256 agcount=4, agsize=91571089 blks = sectsz=512 attr=2 data = bsize=4096 blocks=366284356, imaxpct=5 = sunit=0 swidth=0 blks naming =version 2 bsize=4096 ascii-ci=0 log =internal bsize=4096 blocks=16384, version=2 = sectsz=512 sunit=0 blks, lazy-count=1 realtime =none extsz=4096 blocks=0, rtextents=0 xfs_db -r /dev/md0 xfs_db> frag actual 31835, ideal 31203, fragmentation factor 1.99% xfs_db> freesp from to extents blocks pct 1 1 534 534 0.00 2 3 619 1512 0.00 4 7 848 4596 0.00 8 15 2097 25917 0.03 16 31 697 15674 0.02 32 63 450 20410 0.02 64 127 655 60299 0.06 128 255 646 117567 0.12 256 511 630 237982 0.23 512 1023 221 155404 0.15 1024 2047 67 100821 0.10 2048 4095 51 146798 0.14 4096 8191 68 397150 0.39 8192 16383 634 9578465 9.45 16384 32767 413 11076585 10.93 32768 65535 321 15378630 15.17 65536 131071 193 17909242 17.67 131072 262143 134 23191822 22.88 262144 524287 56 19911605 19.65 524288 1048575 5 3023742 2.98 hdparm -Tt /dev/md0 /dev/md0: Timing cached reads: 7168 MB in 2.00 seconds = 3587.15 MB/sec Timing buffered disk reads: 388 MB in 3.00 seconds = 129.25 MB/sec smartctl indicates no errors from short test. msdam synced the drive last week xfs_repair has run without issue new HDHR Primer 2 old HDHR runs MythTv Realtime processing threads = False Enhanced audio buffering = true Fixed 48Hz sound = true Video is VDPAU high (have tried them all) My next step thoughts: Breaking the raid, Make new multiple partition disk image (filestream mount option) Recreate the Raid1 for each partitiion (force inodes from 32 to 64 mode) Add Raids to MythTv via lvm Break Raid Change FileSystem of Raid Recreate Raid Change IO Scheduler to X from Y Anyone have any insights or suggestions? |
Author: | mattbatt [ Mon Dec 22, 2014 10:39 pm ] |
Post subject: | Re: XFS new HD recordings have large number of extents & cho |
RAID 1 is Redundant RAID right? Maybe one of the drives is getting errors and the two just aren't syncing together? I'd try one drive by it'self if you can see if one or both have worn out. |
Author: | thekingofspain [ Mon Dec 22, 2014 11:26 pm ] |
Post subject: | Re: XFS new HD recordings have large number of extents & cho |
Yes raid 1 is just mirroring. Your suggestion is valid nest step. Eventually that path would have happened at some level if either of the two create new raid the options played out, at least on temporary basis until the new raid was created. Ran into another issue today with a show recording fully, however the commercial falling and fix file size are truncating the show length. There seems to be some posts on this current behavior on the mythtv boards, but it just as well could be a disk issue as well. Trying a transcode now to see the show can get past its bad limit show length attribute. |
Author: | thekingofspain [ Sun Jan 11, 2015 12:10 am ] |
Post subject: | Re: XFS new HD recordings have large number of extents & cho |
Bought a new hard drive and was researching how to break the raid and then rejoin it. To my surprise the raid was already broken. The S.M.A.R.T, which was reporting nothing during the freeze ups, kicked in and failed the drive in 5 minutes which auto broke the raid. Adding the new drive the raid was painless and seemed to sync in about half the time as syncing the previous drive trying when looking for sector discrepancies. Extent usage on the raid is still high but no playback issues. Looks like the issue was a slowly failing drive drive that was just was performing slowly prior is it bad sector failure death |
Author: | mattbatt [ Sun Jan 11, 2015 6:29 pm ] |
Post subject: | Re: XFS new HD recordings have large number of extents & cho |
It's a good thing you had that RAID backup. |
Author: | cahlfors [ Sat Jan 31, 2015 2:19 am ] |
Post subject: | Re: XFS new HD recordings have large number of extents & cho |
@thekingofspain: RAID1 sounds like a very, very good idea and I suspect many of us would like to do that as we cherish our databases and recordings. Would you share how you did it? Perhaps in the hints&tips section? Thanks, /Chris |
Author: | thekingofspain [ Thu Feb 19, 2015 1:02 am ] |
Post subject: | Re: XFS new HD recordings have large number of extents & cho |
I just followed the mythtv wiki on raids like 5 years ago... https://www.mythtv.org/wiki/RAID |
Page 1 of 1 | All times are UTC - 6 hours |
Powered by phpBB® Forum Software © phpBB Group http://www.phpbb.com/ |