August 18, 201015 yr I have a small unRAID server up and running with 3 1TB disks. I previously had 1 tb parity and 2 500gb data disks, and I just yesterday finished upgrading the data disks for 1 TB drives. Even though the main page shows everything fine, all disks green, parity check done this morning perfectly, no errors, etc. I wanted to browse the shares to see if everything was ok... I am curious as to why, if I'm just reading from a disk (either 1 or 2), why does the read counter for the parity drive go up??? Why would it need to read from there if this has supposedly no errors?
August 18, 201015 yr How many reads over what period of time? Querying SMART info or periodic OS checking of drive status could be doing it. Also, when you write to the array, it reads and writes to parity.
August 19, 201015 yr Author Well no, just when I was browsing the files. For example I just checked, about an hour after I wrote the post, and the disks as spinned down, and it has the same just 3 reads it did when I had reset the couters/stats.... I do know it writes every time you write to the array of course, but this is just reading (and browsing, so not actually accessing file for their timestamp to be modified). I just tried browsing some more, and actually opening some files for checking the timestamp thing, and the values did not change... Strange...
August 19, 201015 yr Author Ok, new update: When I browse files that are on DISK1, nothing strange happens... just get reads incremented on DISK1 stat... Now when I browse DISK2, that's when I see Reads being incremented in DISK2, but also on the Parity Disk, and I'm seeing a couple of WRITES on both the Parity Disk and DISK2.... Now what? Like I said, all disks are green, and the parity check done this morning ran fine...
August 19, 201015 yr unRAID mounts disks with noatime and nodiratime, so there is no update of the last access timestamp.
August 19, 201015 yr Did it spindown? I believe changing spin-status from up to down will cause read/writes to be incremented.
August 20, 201015 yr Author Well, yes it did, I mean, Idid the first "test", then the next when I wrote before was like an hour after that, so of course the disks had spun down. I still don't see why spinups would cause there to be writes (or even if it did, why it would keep writing -and to both disks- after it was already spinning constantly...)
August 20, 201015 yr Author Ok, another update... It seams this behaviour of writing to both the data disk and the parity disk, only occurs while browing the shares and folders... that is, say I map the share like so: \\Tower\disk2 in a windows explorer app, I start browsing through the folders, and I get the writes on both disk2 and the parity drive... BUT if you actually open a file (pdf, movie, music, etc) then it will only read from disk2, and I get no reads to the parity drive, and no writes to either the parity drive or disk2... Again, this only happens when you go to disk2, it does not happen on disk 1 Weird or what...? Any dev?
August 20, 201015 yr have you tried: - stop/restart the array - reboot It could be that disk2 somehow mounted w/o the noatime, nodiratime parameters.
August 20, 201015 yr Author have you tried: - stop/restart the array - reboot It could be that disk2 somehow mounted w/o the noatime, nodiratime parameters. Hmm.. well I don't think I rebooted the server after it rebuilt disk2 (upgraded smaller disk), but I do have this in syslog: Aug 18 00:03:33 Tower emhttp: shcmd (13): set -o pipefail ; mount -t reiserfs -o noacl,nouser_xattr,noatime,nodiratime /dev/md2 /mnt/disk2 2>&1 | logger Aug 18 00:03:33 Tower emhttp: shcmd (14): set -o pipefail ; mount -t reiserfs -o noacl,nouser_xattr,noatime,nodiratime /dev/md1 /mnt/disk1 2>&1 | logger That is BEFORE it started the "rebuild" process... I'll reboot the server now and see what happens...
August 20, 201015 yr have you tried: - stop/restart the array - reboot It could be that disk2 somehow mounted w/o the noatime, nodiratime parameters. You could just type mount and see how it was mounted.
August 20, 201015 yr Author Ok, saw it late... I didn't actually reboot, but rather tried just stopping/starting first, and it seems to have stopped.... I did see a couple of writes, but that's just because dring browsing I was dumb enough to go into a pictures folder and that mus have been when the thumbs.db files are created... Steering clear of picture files/folders I see no writes... This is the log after stop/start, seem everything would be ok, right?: Aug 20 00:17:38 Tower emhttp: shcmd (22): /etc/rc.d/rc.samba stop | logger Aug 20 00:17:38 Tower emhttp: shcmd (23): /etc/rc.d/rc.nfsd stop | logger Aug 20 00:17:39 Tower emhttp: Spinning up all drives... Aug 20 00:17:39 Tower emhttp: shcmd (24): sync Aug 20 00:17:39 Tower kernel: mdcmd (17559): spinup 0 Aug 20 00:17:39 Tower kernel: mdcmd (17560): spinup 1 Aug 20 00:17:39 Tower kernel: mdcmd (17561): spinup 2 Aug 20 00:17:47 Tower emhttp: shcmd (25): umount /mnt/user >/dev/null 2>&1 Aug 20 00:17:47 Tower emhttp: shcmd (26): rmdir /mnt/user >/dev/null 2>&1 Aug 20 00:17:47 Tower emhttp: shcmd (27): umount /mnt/disk1 >/dev/null 2>&1 Aug 20 00:17:47 Tower emhttp: shcmd (28): rmdir /mnt/disk1 >/dev/null 2>&1 Aug 20 00:17:47 Tower emhttp: shcmd (29): umount /mnt/disk2 >/dev/null 2>&1 Aug 20 00:17:47 Tower emhttp: shcmd (30): rmdir /mnt/disk2 >/dev/null 2>&1 Aug 20 00:17:47 Tower kernel: mdcmd (17563): stop Aug 20 00:17:47 Tower kernel: md1: stopping Aug 20 00:17:47 Tower kernel: md2: stopping Aug 20 00:17:48 Tower emhttp: shcmd (31): rm /etc/samba/smb-shares.conf >/dev/null 2>&1 Aug 20 00:17:48 Tower emhttp: shcmd (32): cp /etc/exports- /etc/exports Aug 20 00:17:48 Tower emhttp: shcmd (33): /etc/rc.d/rc.samba start | logger Aug 20 00:17:48 Tower logger: Starting Samba: /usr/sbin/nmbd -D Aug 20 00:17:48 Tower logger: /usr/sbin/smbd -D Aug 20 00:17:48 Tower emhttp: shcmd (34): /etc/rc.d/rc.nfsd start | logger Aug 20 00:17:58 Tower emhttp: shcmd (35): /usr/local/sbin/set_ncq sda 1 >/dev/null Aug 20 00:17:58 Tower emhttp: shcmd (36): /usr/local/sbin/set_ncq sdc 1 >/dev/null Aug 20 00:17:58 Tower emhttp: shcmd (37): /usr/local/sbin/set_ncq sdb 1 >/dev/null Aug 20 00:17:58 Tower kernel: mdcmd (17567): start STOPPED Aug 20 00:17:58 Tower kernel: unraid: allocating 18360K for 1280 stripes (3 disks) Aug 20 00:17:58 Tower kernel: md1: running, size: 976762552 blocks Aug 20 00:17:58 Tower kernel: md2: running, size: 976762552 blocks Aug 20 00:17:58 Tower emhttp: shcmd (38): udevadm settle Aug 20 00:17:58 Tower emhttp: shcmd (39): mkdir /mnt/disk2 Aug 20 00:17:58 Tower emhttp: shcmd (39): mkdir /mnt/disk1 Aug 20 00:17:58 Tower kernel: mdcmd (17569): check Aug 20 00:17:58 Tower kernel: md: recovery thread woken up ... Aug 20 00:17:58 Tower emhttp: shcmd (40): set -o pipefail ; mount -t reiserfs -o noacl,nouser_xattr,noatime,nodiratime /dev/md2 /mnt/disk2 2>&1 | logger Aug 20 00:17:58 Tower emhttp: shcmd (41): set -o pipefail ; mount -t reiserfs -o noacl,nouser_xattr,noatime,nodiratime /dev/md1 /mnt/disk1 2>&1 | logger Aug 20 00:17:58 Tower kernel: md: recovery thread has nothing to resync Aug 20 00:17:58 Tower kernel: REISERFS (device md1): found reiserfs format "3.6" with standard journal Aug 20 00:17:58 Tower kernel: REISERFS (device md1): using ordered data mode Aug 20 00:17:58 Tower kernel: REISERFS (device md2): found reiserfs format "3.6" with standard journal Aug 20 00:17:58 Tower kernel: REISERFS (device md2): using ordered data mode Aug 20 00:17:58 Tower kernel: REISERFS (device md1): journal params: device md1, size 8192, journal first block 18, max trans len 1024, max batch 900, max commit age 30, max trans age 30 Aug 20 00:17:58 Tower kernel: REISERFS (device md1): checking transaction log (md1) Aug 20 00:17:58 Tower kernel: REISERFS (device md2): journal params: device md2, size 8192, journal first block 18, max trans len 1024, max batch 900, max commit age 30, max trans age 30 Aug 20 00:17:58 Tower kernel: REISERFS (device md2): checking transaction log (md2) Aug 20 00:17:59 Tower kernel: REISERFS (device md1): Using r5 hash to sort names Aug 20 00:17:59 Tower kernel: REISERFS (device md2): Using r5 hash to sort names Aug 20 00:17:59 Tower emhttp: shcmd (43): rm /etc/samba/smb-shares.conf >/dev/null 2>&1 Aug 20 00:17:59 Tower emhttp: _shcmd: shcmd (43): exit status: 1 Aug 20 00:17:59 Tower emhttp: shcmd (44): cp /etc/exports- /etc/exports Aug 20 00:17:59 Tower emhttp: shcmd (45): mkdir /mnt/user Aug 20 00:17:59 Tower emhttp: shcmd (46): /usr/local/sbin/shfs /mnt/user -o noatime,big_writes,allow_other,default_permissions Aug 20 00:18:00 Tower emhttp: shcmd (47): killall -HUP smbd Aug 20 00:18:00 Tower emhttp: shcmd (48): /etc/rc.d/rc.nfsd restart | logger Aug 20 00:18:11 Tower kernel: mdcmd (17574): clear Aug 20 00:23:51 Tower kernel: mdcmd (17637): clear
Archived
This topic is now archived and is closed to further replies.