Re: [SlimDevices: Unix] SSD question
FYI, fsck time for the 3T ext4 media drive, now 37% full: Code: # time fsck -f /dev/sdc1 fsck from util-linux 2.19.1 e2fsck 1.41.14 (22-Dec-2010) Pass 1: Checking inodes, blocks, and sizes Pass 2: Checking directory structure Pass 3: Checking directory connectivity Pass 4: Checking reference counts Pass 5: Checking group summary information Media: 9681/45787136 files (4.4% non-contiguous), 270791081/732566385 blocks real1m27.183s user0m47.691s sys 0m3.840s So..about a minute and a half...which seems very tolerable to me. I'll test again after I have the video data on the drive which ought to bring it to about 80% full. -- gharris999 gharris999's Profile: http://forums.slimdevices.com/member.php?userid=115 View this thread: http://forums.slimdevices.com/showthread.php?t=93717 ___ unix mailing list unix@lists.slimdevices.com http://lists.slimdevices.com/mailman/listinfo/unix
Re: [SlimDevices: Unix] SSD question
epoch: thanks for that. So, in a nutshell: Don't worry about SSD wear. Just use the damn drive, damnit. OK, I'll go with that. With the system occupying only 10% of /dev/sda1, I suppose that gives the drive firmware a lot of latitude to spread the wear around. FYI: from grub menu to login prompt: 11 seconds on this machine. Not too shabby. Mobo is a Asus E35M1-I mini-itx board. AMD dual core cpu, 6 (six!) SATA 3 ports. Eventually, this will be my NAS box: 3TB media drive serving audio to LMS and video to minidlna. I'll add 4 2TB drives in a soft-raid array for general storage / data backup. The challenge there: I'll be hitting the raid array only occasionally so ideally, I'd like those drives to idle and spin down. I don't know yet if mdadm can deal with that, though. Re the media drive: with just static data, why journal at all? -- gharris999 gharris999's Profile: http://forums.slimdevices.com/member.php?userid=115 View this thread: http://forums.slimdevices.com/showthread.php?t=93717 ___ unix mailing list unix@lists.slimdevices.com http://lists.slimdevices.com/mailman/listinfo/unix
Re: [SlimDevices: Unix] SSD question
What does the drive manufacturer say about read/write cycles or other similar specs? Is it different from a disk? Personally, I wouldn't spend the money for an SSD until the Just use the damn drive rule applies. -- Mark Miksis Mark Miksis's Profile: http://forums.slimdevices.com/member.php?userid=529 View this thread: http://forums.slimdevices.com/showthread.php?t=93717 ___ unix mailing list unix@lists.slimdevices.com http://lists.slimdevices.com/mailman/listinfo/unix
Re: [SlimDevices: Unix] SSD question
The real question would be why would you pay the extra and use an SSD? My one and only reason is that HD add to the acoustic noise level. A smaller but not less important reason would be power consumption. In my LMS-machine OS sits on the SSD (Samsung SSD PB22-J MLC), without swap, and /tmp sitting ram. Code: devpts /dev/pts devpts defaults0 0 shm/dev/shm tmpfs nodev,nosuid,size=1G0 0 none /tmp tmpfs nodev,nosuid,size=2G,mode=1777 0 0 #LABEL=swap swap swap defaults0 0 LABEL=root / ext4 defaults,noatime,nodiratime,discard 0 1 LABEL=boot /boot ext4 defaults,noatime,nodiratime,discard 0 3 LABEL=home /home ext4 defaults,noatime,nodiratime,discard 0 2 The media files are on a RAID5 (HD) based LVM. I did experiment with the log files on a tmpfs as well but ran in a number weird boot problems so stopped doing that. Ofcourse /opt/logitechmediaserver sits on the SSD so DB activity is fast and noiseless. The audio files are located on a RAID5 (HD) based lvm set which can spin-down if not needed. I would not worry about the SSD. You are not planning to write GBytes of new data daily (are you?) only once most probably. The SSD most likely will outlive the economic value of your system. -- th00ht Ripping: 'EAC (free and great)' (http://www.exactaudiocopy.de/) Squeezebox Server 7.7.1 - r33735 / Arch64 Squeezebox Classic, Quad 303 (where can I get new caps?) + two Quad ESL 63 Squeezebox Touch, Denon AVR 1912 (Pure Direct) + two CM8 + two CM1 + CMC Squeezebox Radio White 'last.fm' (http://www.last.fm/user/th00ht), 'SoS' (http://www.bowers-wilkins.com/Society_of_Sound/Society_of_Sound/Music/download-manager.html), 'linn' ( http://www.linn.co.uk/music) th00ht's Profile: http://forums.slimdevices.com/member.php?userid=15656 View this thread: http://forums.slimdevices.com/showthread.php?t=93717 ___ unix mailing list unix@lists.slimdevices.com http://lists.slimdevices.com/mailman/listinfo/unix
Re: [SlimDevices: Unix] SSD question
Very, very good IOs and throughput vote in favor of SSDs. In my case, the SSD on a spare SATA 2 port on the board was much faster than the original 10k SAS drive. Consequences: - The freed SAS port allowed me to expand a storage array - The system is a bit faster, but I don't really care for that because the machine is mostly a host to VMs (and it boots 3 or 4 times a year) - My VMs are much, much faster. It's all going so well that I could virtualize part of the network (switches, firewalls) with no penalty compared to my previous setup with (simpler) physical-only network. I also expected good reliability. I chose a trusted model from a trusted brand. I doubt all SSDs are engineered or built equal. To be clear: nothing of this says you want an SSD to run LMS. But if you're using virtual machines and expect consistent response time, you might consider the option. gharris999;692007 Wrote: Eventually, this will be my NAS box: 3TB media drive serving audio to LMS and video to minidlna. I'll add 4 2TB drives in a soft-raid array for general storage / data backup. The challenge there: I'll be hitting the raid array only occasionally so ideally, I'd like those drives to idle and spin down. I don't know yet if mdadm can deal with that, though. Re the media drive: with just static data, why journal at all? I'd say because when you crash or simply hit regular maintenance time, disk checking/repair would take an immense time without journal. In the case of a crash I am sure. For regular maintenance I am not certain there is a difference. -- epoch1970 Daily dose delivered by: 3 SB Classic, 1 SB Boom iPeng (iPhone + iPad) Squeezebox Server 7.6 (Debian 5.0) with plugins: MusicIP Server Power Control by Gordon Harris WeatherTime by Martin Rehfeld IRBlaster by Gwendesign (Felix) Find cover art by bpa BBC iPlayer, SwitchPlayer by Triode PowerSave by Jason Holtzapple Song Info, Song Lyrics by Erland Isaksson SaverSwitcher, ContextMenu by Peter Watkins Just Covers by Tom Kalmijn. epoch1970's Profile: http://forums.slimdevices.com/member.php?userid=16711 View this thread: http://forums.slimdevices.com/showthread.php?t=93717 ___ unix mailing list unix@lists.slimdevices.com http://lists.slimdevices.com/mailman/listinfo/unix
Re: [SlimDevices: Unix] SSD question
epoch1970;692032 Wrote: I'd say because when you crash or simply hit regular maintenance time, disk checking/repair would take an immense time without journal. In the case of a crash I am sure. For regular maintenance I am not certain there is a difference. (w/ journal, checking a 6TB volume takes *hours*) EDIT: Unless you mount the drives as read-only. Could it be an option? My current production LMS box has the media files on the same model disk, ext4 with journal and no special tweaking of parameters. With this new box, I thought I'd mess with inodes, journal such to try to squeeze a little more disk space and performance from the drive and reduce system overhead. I'll populate the drive over the next day or two, simulate a crash and precipitate a boot time fsck on it and report back the results. If, as you suspect, it takes many hours, I'll likely blow away the partition and start over, keeping the journal. Mounting RO would be a little cumbersome for my setup, given how frequently I push new music and videos up to the system. -- gharris999 gharris999's Profile: http://forums.slimdevices.com/member.php?userid=115 View this thread: http://forums.slimdevices.com/showthread.php?t=93717 ___ unix mailing list unix@lists.slimdevices.com http://lists.slimdevices.com/mailman/listinfo/unix
Re: [SlimDevices: Unix] SSD question
@epoch: I'm a noob when it comes to interpreting smart data. Does the: Code: 233 Media_Wearout_Indicator 0x0032 099 099 000Old_age Always - 0 ..indicate that your drive is 1% worn out? This is what my drive is reporting for that metric: Code: 233 Media_Wearout_Indicator 0x0032 100 100 000Old_age Always - 0 -- gharris999 gharris999's Profile: http://forums.slimdevices.com/member.php?userid=115 View this thread: http://forums.slimdevices.com/showthread.php?t=93717 ___ unix mailing list unix@lists.slimdevices.com http://lists.slimdevices.com/mailman/listinfo/unix
Re: [SlimDevices: Unix] SSD question
Mark Miksis;692008 Wrote: What does the drive manufacturer say about read/write cycles or other similar specs? Is it different from a disk? Personally, I wouldn't spend the money for an SSD until the Just use the damn drive rule applies.Normally, I'd be with you on this. But I happened to have had a small windfall burning a hole in my pocket and with Thai-flooded rotating disk prices being so high...well, let's just say that this was one of my relatively rare impulse purchases. As to specs: http://www.intel.com/content/www/us/en/solid-state-drives/ssd-520-specification.html ..and.. http://www.tomshardware.com/reviews/ssd-520-sandforce-review-benchmark,3124-11.html Paraphrasing from that Tom's article, this drive will be worn out in about 5 years if I write 7gigs of incompressible data to the drive every day. ..Which is probably unlikely. More likely is that 75 year figure...by which time I'll probably be running in a VM myself and LMS will be nothing more than a optional module installed to my virtual neo-cortex. So...back to just use the damn drive. -- gharris999 gharris999's Profile: http://forums.slimdevices.com/member.php?userid=115 View this thread: http://forums.slimdevices.com/showthread.php?t=93717 ___ unix mailing list unix@lists.slimdevices.com http://lists.slimdevices.com/mailman/listinfo/unix
Re: [SlimDevices: Unix] SSD question
gharris999;692077 Wrote: @epoch: I'm a noob when it comes to interpreting smart data. Does the: Code: 233 Media_Wearout_Indicator 0x0032 099 099 000Old_age Always - 0 ..indicate that your drive is 1% worn out? This is what my drive is reporting for that metric: Code: 233 Media_Wearout_Indicator 0x0032 100 100 000Old_age Always - 0 I commend your sharp eye. I googled a bit and read that indeed this seems to be some kind of health point indicator. A new dirve starts at 100. So mine is 99/100 after 13.000 hours of operation, roughly 1.5 year. So it should be good to go for a century ??? This paper has the authoritative information I think: ftp://download.intel.com/design/flash/NAND/325551.pdf (I filter FTP so I didn't read it, but I think it will confirm what I've seen on the web.) In all fairness, my system swaps very little, as there is a lot of RAM available. On the other hand, this drive supports in fact the continuous r/w, logging#8230; operation of the host + 4 other linux VMs, + 1 m0n0wall (embedded bsd) firewall. And a few more from time to time. -- epoch1970 Daily dose delivered by: 3 SB Classic, 1 SB Boom iPeng (iPhone + iPad) Squeezebox Server 7.6 (Debian 5.0) with plugins: MusicIP Server Power Control by Gordon Harris WeatherTime by Martin Rehfeld IRBlaster by Gwendesign (Felix) Find cover art by bpa BBC iPlayer, SwitchPlayer by Triode PowerSave by Jason Holtzapple Song Info, Song Lyrics by Erland Isaksson SaverSwitcher, ContextMenu by Peter Watkins Just Covers by Tom Kalmijn. epoch1970's Profile: http://forums.slimdevices.com/member.php?userid=16711 View this thread: http://forums.slimdevices.com/showthread.php?t=93717 ___ unix mailing list unix@lists.slimdevices.com http://lists.slimdevices.com/mailman/listinfo/unix
Re: [SlimDevices: Unix] SSD question
gharris999;692073 Wrote: My current production LMS box has the media files on the same model disk, ext4 with journal and no special tweaking of parameters. With this new box, I thought I'd mess with inodes, journal such to try to squeeze a little more disk space and performance from the drive and reduce system overhead. I'll populate the drive over the next day or two, simulate a crash and precipitate a boot time fsck on it and report back the results. If, as you suspect, it takes many hours, I'll likely blow away the partition and start over, keeping the journal. Mounting RO would be a little cumbersome for my setup, given how frequently I push new music and videos up to the system. My server holds a 6TiB and a 10TiB raid5 array. Around 40% full each. Slow SATA drives harnessed by a hw raid-5 card. Dual CPU dual core, still a relatively fast machine. When I reboot I sometimes see the dreaded message: This fs has spent over 180 days without check, checking now. And then, reaching the boot prompt takes the better part of the day About blocksize/inode count, you could probably opt for a smaller block size than the default used by -T largefiles. Compared to a regular filesystem, data stores tend to use simpler directory structure, hold little to none soft/hard links, are not used for scratch/spooling. So there is less chance you'd run out of inodes before running out of space, even with a smaller block size. How much of a gain this indeed represents depends on what proportion of very small files the big fs would hold. In an ocean of GBs, what does, say, 100.000 wasteful 32k files ultimately represent ? -- epoch1970 Daily dose delivered by: 3 SB Classic, 1 SB Boom iPeng (iPhone + iPad) Squeezebox Server 7.6 (Debian 5.0) with plugins: MusicIP Server Power Control by Gordon Harris WeatherTime by Martin Rehfeld IRBlaster by Gwendesign (Felix) Find cover art by bpa BBC iPlayer, SwitchPlayer by Triode PowerSave by Jason Holtzapple Song Info, Song Lyrics by Erland Isaksson SaverSwitcher, ContextMenu by Peter Watkins Just Covers by Tom Kalmijn. epoch1970's Profile: http://forums.slimdevices.com/member.php?userid=16711 View this thread: http://forums.slimdevices.com/showthread.php?t=93717 ___ unix mailing list unix@lists.slimdevices.com http://lists.slimdevices.com/mailman/listinfo/unix
Re: [SlimDevices: Unix] SSD question
epoch1970, fletch, th00ht: Thanks for all your input here. Once I'm done tweaking this system, I'll post my mount options for the SSD for anyone else interested in this in the future. For now, I'm mounting the SSD in fstab with: Code: UUID=d017202e-5a71-4480-aa57-edbfa6531ec0 / ext4discard,noatime,nodiratime,errors=remount-ro0 1 In the mean time, broadening the topic of this thread, here are a couple of other observations and, again, please chime in with your wisdom: I git-ed the latest code for gptfdisk and built it and used it to partition the Seagate ST3000DM001 3T drive. I then created the file system with: Code: mkfs.ext4 -b 4096 -O ^has_journal,extent,huge_file,flex_bg,uninit_bg,dir_nlink,extra_isize -m 0 -i 65536 -L Media /dev/sdb1 tune2fs -i 0 -c 0 /dev/sdb1 I'm operating a little in the dark with that one. I think what I've accomplished with that was to: - Disable journaling - Set the reserved blocks to 0 - Dial down the number of inodes - Disable automatic fsck I'm mounting the drive in fstab via: Code: UUID=877a1c84-2646-4433-8f74-7407348582e8 /mnt/media ext4defaults,data=writeback,noatime,nodiratime,errors=remount-ro 0 0 Again, I'm operating a little above my level of expertise, so feel free to amend my errant ways. I also downloaded the latest rsync code and built the latest version that supports the --preallocate option. I have a long term quest to write a bash script that will allow me to never have to master rsync syntax and instead use windows' robocopy syntax (which is indelibly etched into my neurons). Robocopy, attempting to fight file fragmentation, preallocates disk space when it copies files. I thought I'd use this new version of rsync to duplicate that behavior. Oddly, the preallocation only seemed to work with the first directory copied. With subsequent directories, rsync seems to revert to it's default behavior. Copying from one ST3000DM001 to another (and with both mounted using the same options) I'm seeing copy speeds top out at about 40 mb / second. That's when copying 200-300mb flacs. That still seems a little slow to me given that the mobo's SATA interface and both drives are SATA III / 6.0 Gb/s. This is the rsync command I'm using: Code: rsync --links --perms --times --group --owner --devices --specials --stats --progress --verbose --preallocate --recursive --include=*/ --include=* --exclude=* /mnt/media/Music/ /mnt/medianew/Music -- gharris999 gharris999's Profile: http://forums.slimdevices.com/member.php?userid=115 View this thread: http://forums.slimdevices.com/showthread.php?t=93717 ___ unix mailing list unix@lists.slimdevices.com http://lists.slimdevices.com/mailman/listinfo/unix
[SlimDevices: Unix] SSD question
Ola, linux wonks! For the first time, I'm building a slimserver box using a SSD as the boot drive. I'd appreciate it if folks would weigh-in on the appropriate system settings and optimizations that are advisable with such a setup. The box is currently: - Using an Intel 520 60gb SSD with the newest firmware as the boot drive - Running Ubuntu 11.10 server, headless - Mounting the ext4 / with discard,noatime,nodiratime - Not mounting the swap partition (won't be hibernating, only suspending; system has 8gigs ram) - Using a Seagate ST3000DM001 3T drive (ext4) for the audio video files Specifically, what is the current wisdom regarding SSD read/write wear? Do new wear leveling features in SSDs make this no longer a concern? Alternately, beyond what I've already done, would folks recommend: - Moving /tmp /var/tmp /log to tmpfss? - Or just moving /var/squeezeboxserver to a tmpfs? - Use boot, halt and reboot scripts to copy /var /log from and to rotating disks to and from tempfss? Other considerations? -- gharris999 gharris999's Profile: http://forums.slimdevices.com/member.php?userid=115 View this thread: http://forums.slimdevices.com/showthread.php?t=93717 ___ unix mailing list unix@lists.slimdevices.com http://lists.slimdevices.com/mailman/listinfo/unix
Re: [SlimDevices: Unix] SSD question
I've been using an Intel X25-M 80 Go PostVille as the boot drive on my main server (Dell PE 2900) for 1 1/2 years. I have no issue at all. I didn't tweak the filesystem or whatnot, it just works. It runs the OS, there is a swap partition on it, and the system drives of all my most important/used VMs are stored on it (many running 24/7 as the server does.) Below some system info reported from 4 commands, to see for yourself. Some comments: i) when I wrote comments in fstab years ago I lied: the system drive is not call sda, it is indeed sdc ii) /proc/cmdline contains the current grub boot options. How do you like that ? iii) as said, a plain ext3 drive. Last partition is used for /home/vms where my VMs HDD images are. iv) In the very thorough smartctl trace you'll read a huge value under 228 Power-off_Retract_Count. After worrying (given the funky way the drive is powered, via the unused floppy power feed) and googling, it appears that this report is erroneous due to an old smartmontool version. The drive is indeed O.K. Code: $ grep -C8 Intel /etc/fstab # /etc/fstab: static file system information. # # file system mount point type options dump pass proc/proc procdefaults0 0 ## ## System disk (SATA Port 0 - Located in floppy drive slot - Powered via floppy_pwr, 3.3V only) ## Intel SSD 80G - Jul 25 2010 ## A Dell diags partition is under sda1. ## # sda6 LABEL=SYS-100725 /ext3errors=remount-ro 0 1 # sda2 LABEL=BOOT-100725 /boot ext3defaults0 2 # sda5 LABEL=SWAP1-20G none swapsw 0 0 $ cat /proc/cmdline root=LABEL=SYS-100725 ro selinux=0 apparmor=0 security= console=tty0 console=ttyS1,57600n8 quiet noresume $ sudo fdisk -l /dev/sdc Disk /dev/sdc: 80.0 GB, 80026361856 bytes 255 heads, 63 sectors/track, 9729 cylinders Units = cylinders of 16065 * 512 = 8225280 bytes Disk identifier: 0x373f9603 Device Boot Start End Blocks Id System /dev/sdc1 1 5 40131 de Dell Utility /dev/sdc2 6 30 200812+ 83 Linux /dev/sdc3 31972977907217+ 5 Extended /dev/sdc5 31246319543041 82 Linux swap / Solaris /dev/sdc62464972958364113+ 83 Linux $ sudo smartctl -a /dev/sdc smartctl 5.40 2010-07-12 r3124 [x86_64-unknown-linux-gnu] (local build) Copyright (C) 2002-10 by Bruce Allen, http://smartmontools.sourceforge.net === START OF INFORMATION SECTION === Model Family: Intel X25-M SSD Device Model: INTEL SSDSA2M080G2GC Serial Number:CVPO017001SP080JGN Firmware Version: 2CV102HD User Capacity:80,026,361,856 bytes Device is:In smartctl database [for details use: -P show] ATA Version is: 7 ATA Standard is: ATA/ATAPI-7 T13 1532D revision 1 Local Time is:Tue Feb 21 23:02:27 2012 CET SMART support is: Available - device has SMART capability. SMART support is: Enabled === START OF READ SMART DATA SECTION === SMART overall-health self-assessment test result: PASSED General SMART Values: Offline data collection status: (0x00) Offline data collection activity was never started. Auto Offline Data Collection: Disabled. Self-test execution status: ( 0) The previous self-test routine completed without error or no self-test has ever been run. Total time to complete Offline data collection: ( 1) seconds. Offline data collection capabilities: (0x75) SMART execute Offline immediate. No Auto Offline data collection support. Abort Offline collection upon new command. No Offline surface scan supported. Self-test supported. Conveyance Self-test supported. Selective Self-test supported. SMART capabilities:(0x0003) Saves SMART data before entering power-saving mode. Supports SMART auto save timer. Error logging capability:(0x01) Error logging supported. General Purpose Logging supported. Short self-test routine recommended polling time: ( 1) minutes. Extended self-test routine recommended polling time: ( 1) minutes. Conveyance self-test routine recommended polling time: ( 1) minutes. SMART