[zfs-discuss] Snapshot access from non-global zones
Hello I'm curious what is the best (if any) way to access ZFS snapshots from the non-global zones? I have a common ZFS file system (on Solaris Express b116 for ex.) in a global zone, mounted as lofs to several non-global zones. In each zone I can access all files with no problem, but I'm unable to the snapshots. I can get the list of snapshots, but not the content of individual snapshot, all commands fails with "Not owner". There is one trick however - if snapshot was created *before* the filesystem was lofs mounted inside of a zone these snapshots are perfectly accessible. But if the snapshots were created after the mount - they are not accessible from inside of a zone. So this is correct behavior or it's bug, any workarounds? Thanks in advance for all comments. Regards, Mike ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] j4200 drive carriers
Hello > 1) Dual IO module option > 2) Multipath support > 3) Zone support [multi host connecting to same JBOD or same set of JBOD's > connected in series. ] This sounds interesting - where I can read more about connecting two hosts to same J4200 etc? Thanks Mike ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] zfs send / zfs receive hanging
Hi > It would be also nice to be able to specify the zpool version during pool > creation. E.g. If I have a newer machine and I want to move data to an older > one, I should be able to specify the pool version, otherwise it's a one-way > street. zpool create -o version=xx ... Mike ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] zfs list improvements?
Hello > This seems like a reasonable proposal to enhance zfs list. But it would > also be good to add as few new options to zfs list as possible. So it > probably makes sense to add at most one of these new options. Or > perhaps add an optional depth argument to the -r option instead? > > As you point out, the -c option is user friendly while the -depth (or > maybe -d) option is more general. There have been several requests for > the -c option. Would anyone prefer the -depth option? In what cases > would this be used? I'd have a few more proposals how to improve zfs list if they don't contravene the concept of zfs list command. Currently zfs list returns error "operation not applicable to datasets of this type" if you try to list for ex.: "zfs list -t snapshot file/system" returns above error while it could return what you actually asked - the list of all snapshots of "file/system". Similar case is if you try "zfs list file/sys...@snapshot" - can zfs be more smart to return the snapshot instead of error message if dataset name contains "@" in its name? Other thing is zfs list performance... even if you want to get the list of snapshots with no other properties "zfs list -oname -t snapshot -r file/system" it still takes quite long time if there are hundreds of snapshots, while "ls /file/system/.zfs/snapshot" returns immediately. Can this also be improved somehow please? Thanks Mike ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] 'zfs recv' is very slow
Hello > Yah, the incrementals are from a 30TB volume, with about 1TB used. > Watching iostat on each side during the incremental sends, the sender > side is hardly doing anything, maybe 50iops read, and that could be > from other machines accessing it, really light load. > The receiving side however, for about 3 minutes it is peaking around > 1500 iops reads, and no writes. Have you tries truss on both sides? From my experiments I found that sending side on beginning of the transfer mostly sleeps while receiving lists all available snapshots on the syncing file system. So if you have a lot of snapshots on receiving side (as in my case) the process will take long time sending no data but listing the snapshots. The worst case is if you use recursive sync of hundreds of file system with hundreds of snapshots on each. I'm sure this must be optimized somehow otherwise it's almost useless in practice. Regards Mike ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] How to compile mbuffer
Hello > i've been following the "[zfs-discuss] 'zfs recv' is very slow" thread > and i believe i have the same issue; we get ~10MB/sec sending large > incrimental data sets using zfs send | ssh | zfs recv. I'd like to try > mbuffer. > > We're running Solaris Express Developers Edition (SunOS murray 5.11 > snv_79a i86pc i386 i86pc). I found the download page > http://www.maier-komor.de/mbuffer.html and i have the source files on > Murray. > > How do i compile mbuffer for our system, and what syntax to i use to > invoke it within the zfs send recv? > > Any help appreciated! I used compile it this way: 1) wget http://www.maier-komor.de/software/mbuffer/mbuffer-20081113.tgz 2) gtar -xzvf mbuffer-20081113.tgz 3) cd mbuffer-20081113 4) ./configure --prefix=/usr/local --disable-debug CFLAGS="-O" MAKE=gmake If you are on 64bit system you may want to compile 64bit version: ./configure --prefix=/usr/local --disable-debug CFLAGS="-O -m64" MAKE=gmake 5) gmake && gmake install 6) /usr/local/bin/mbuffer -V Regards Mike ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
[zfs-discuss] ZFS performance
Hello list, I have a system with 2x 1.8 GHz AMD CPUs, 4G of ECC RAM, 7T RAID-Z pool on Areca controller with about 400 file systems on OpenSolaris snv_101. The problem is that it takes VERY long to take or delete snapshot and sync incremental snapshots to backup system. System load is quite low I'd say, CPU is 98% idle: load average: 0.09, 0.13, 0.26 IOPs are low as well: capacity operationsbandwidth pool used avail read write read write -- - - - - - - data5.62T 2.32T 98115 7.72M 2.35M data5.62T 2.32T547227 47.9M 864K data5.62T 2.32T204 58 15.9M 616K data5.62T 2.32T 4 0 256K 0 data5.62T 2.32T 20 0 399K 0 data5.62T 2.32T 99 47 9.68M 264K data5.62T 2.32T 0 11 6.93K 38.1K data5.62T 2.32T 0455506 1.90M data5.62T 2.32T250 21 17.0M 420K data5.62T 2.32T150235 10.7M 1.34M data5.62T 2.32T305 0 16.0M 0 data5.62T 2.32T137 3.42K 12.9M 16.8M data5.62T 2.32T107 0 13.2M 0 data5.62T 2.32T 56 0 4.97M 0 data5.62T 2.32T200296 23.6M 1.70M mpstat output: CPU minf mjf xcal intr ithr csw icsw migr smtx srw syscl usr sys wt idl 0 160 0 690 1152 568 1133 89 68 1440 25995 8 0 87 1 154 0 108 4424 3241 1388 102 68 1370 24815 7 0 88 CPU minf mjf xcal intr ithr csw icsw migr smtx srw syscl usr sys wt idl 06 0 83 594 365 2860 3130 6160 7 0 93 10 00 524 141 6692 2710 3210 2 0 98 CPU minf mjf xcal intr ithr csw icsw migr smtx srw syscl usr sys wt idl 00 0 55 575 353 2800 1530 4831 6 0 93 10 00 462 142 6103 1750 4501 2 0 97 CPU minf mjf xcal intr ithr csw icsw migr smtx srw syscl usr sys wt idl 00 00 454 210 3230 1910 7630 3 0 97 10 00 288 166 2970 1530 3380 2 0 98 CPU minf mjf xcal intr ithr csw icsw migr smtx srw syscl usr sys wt idl 00 00 398 172 2130 1310 6260 1 0 99 10 00 252 154 2450 1510 2490 0 0 100 CPU minf mjf xcal intr ithr csw icsw migr smtx srw syscl usr sys wt idl 00 00 461 229 2920 1710 5010 3 0 97 10 00 290 149 3394 1210 4020 2 0 98 What can be wrong that ZFS operations like create file system, take/destroy snapshot, (not saying snapshot listing which takes ages) takes minutes to complete. Is there something I can look at which would help to determine where is a bottleneck or what is wrong etc? Thanks in advance for any advice, Mike ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] ZFS snapshot list
Hi > [Default] On Sat, 15 Nov 2008 11:37:50 +0200, Mike Futerko > <[EMAIL PROTECTED]> wrote: > >> Hello >> >> Is there any way to list all snapshots of particular file system >> without listing the snapshots of its children file systems? > > fsnm=tank/fs;zfs list -rt snapshot ${fsnm}|grep "${fsnm}@" > > or even > > fsnm=tank/fs;zfs list -r ${fsnm}|grep "${fsnm}@" Yes, thanks - I know about grep but if you have hundred of thousands of snapshots grep is what I wanted to avoid. In my case full zfs list -rt snapshot take hours, while listing snapshot for individual filesystem is much much quicker :( Regards Mike ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
[zfs-discuss] ZFS snapshot list
Hello Is there any way to list all snapshots of particular file system without listing the snapshots of its children file systems? Thanks, Mike ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] ZFS on Fit-PC Slim?
Hi > Not merely a little pokey it was unacceptably slow and the casing got very > warm. I am guessing it was pushing CPU right to 100% all the time. Took > hours to load and when booting took minutes. Also didn't see an easy way to > disable graphical login so on boot every time it would go to scrambled > graphics and then no easy way to get in to turn off graphics. Trying virtual > console switching didn't work for me. It didn't take me long to realize this > wasn't going to be usable. Well... it's easy to disable graphical login: svcadm disable cde-login I'd also recommend to disable some other unnecessary processes, ex: svcs | egrep 'webco|wbem|avahi|print|font|cde|sendm|name-service-cache|opengl' | awk '{print $3}' | xargs -n1 svcadm disable This should made your system more usable on light hardware. Regards Mike ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] Strange result when syncing between SPARC and x86
Hi >> [EMAIL PROTECTED] ls -la >> /data/zones/testfs/root/etc/services >> lrwxrwxrwx 1 root root 15 Oct 13 14:35 >> /data/zones/testfs/root/etc/services -> >> ./inet/services >> >> [EMAIL PROTECTED] ls -la /data/zones/testfs/root/etc/services >> lrwxrwxrwx 1 root root 15 Oct 13 14:35 >> /data/zones/testfs/root/etc/services -> >> s/teni/.ervices > > Ouch, thats a bad one. > > I downloaded and burnt b101 to dvd for x86 and solaris, i'm gonna install > them tomorrow at work and try moving a pool between them to see what > happens... Would be interesting to know how it'll work if move whole zpool not just sync with send/recv. But I think all will be fine there as is seems the problem is in send/recv part on the file system itself on different architectures. Thanks Mike ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
Re: [zfs-discuss] Strange result when syncing between SPARC and x86
Hi Just checked with snv_99 on x86 (VMware install) - same result :( Regards Mike [EMAIL PROTECTED] wrote: >> Hello >> >> >> Today I've suddenly noticed that symlinks (at least) are corrupted when >> sync ZFS from SPARC to x86 (zfs send | ssh | zfs recv). >> >> Example is: >> >> [EMAIL PROTECTED] ls -la /data/zones/testfs/root/etc/services >> lrwxrwxrwx 1 root root 15 Oct 13 14:35 >> /data/zones/testfs/root/etc/services -> ./inet/services >> >> [EMAIL PROTECTED] ls -la /data/zones/testfs/root/etc/services >> lrwxrwxrwx 1 root root 15 Oct 13 14:35 >> /data/zones/testfs/root/etc/services -> s/teni/.ervices >> >> >> Firstly I thought it's because original FS on SPARC is compressed... so >> I've just synced it locally on same machine and all was OK just >> different FS size since destination was not compressed. >> >> Then I've synced that copy again to x86 but result was same - symlinks >> were corrupted... so it's not compression. >> >> SPARC is snv_85 and x86 snv_82, I haven't got a chance yet to test on >> latest OpenSolaris. >> >> >> Any suggestions? > > Looks like the first 8 bytes aren't "reversed". > > Casper > ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
[zfs-discuss] Strange result when syncing between SPARC and x86
Hello Today I've suddenly noticed that symlinks (at least) are corrupted when sync ZFS from SPARC to x86 (zfs send | ssh | zfs recv). Example is: [EMAIL PROTECTED] ls -la /data/zones/testfs/root/etc/services lrwxrwxrwx 1 root root 15 Oct 13 14:35 /data/zones/testfs/root/etc/services -> ./inet/services [EMAIL PROTECTED] ls -la /data/zones/testfs/root/etc/services lrwxrwxrwx 1 root root 15 Oct 13 14:35 /data/zones/testfs/root/etc/services -> s/teni/.ervices Firstly I thought it's because original FS on SPARC is compressed... so I've just synced it locally on same machine and all was OK just different FS size since destination was not compressed. Then I've synced that copy again to x86 but result was same - symlinks were corrupted... so it's not compression. SPARC is snv_85 and x86 snv_82, I haven't got a chance yet to test on latest OpenSolaris. Any suggestions? Thanks Mike ___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss
[zfs-discuss] Kernel panic on ZFS snapshot destroy
Hello all I have weird problem with a snapshot... when I try to delete it kernel panics. However I can successfully create and then delete other snapshots on same file system. The OS version I noticed it happens was snv_81 so I've upgraded to snv_94 (LU) but it doesn't help. I've attached a screenshot if it may be useful. Any help would be appreciated... Thanks, Mike <>___ zfs-discuss mailing list zfs-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/zfs-discuss