Re: [Vserver] dev setting in interfaces
Stephan Mueller wrote: Hi Herbert, * Herbert Poetzl [EMAIL PROTECTED] [051222 18:10]: if you want the tools to 'create' the IP for your guest, then you have to tell them on which interface it shall be done ... ah, okay. Thank your for putting this straight. One more question: where has the onboot option gone in the new configuration style? I was not able to mark my clients to be run after booting the vserver. echo default /etc/vservers/name/apps/init/mark the vservers-default start script must be in the active runlevel! Cheers + thanks again, ___ Vserver mailing list Vserver@list.linux-vserver.org http://list.linux-vserver.org/mailman/listinfo/vserver ___ Vserver mailing list Vserver@list.linux-vserver.org http://list.linux-vserver.org/mailman/listinfo/vserver
Re: [Vserver] vserver migration 1.2.10 2.0.1
Ok, this is really emberassing to me. I forgot to use --numeric-owner with tar. Though, I don't know why (without --numeric-owner) only the GIDs got messed up and not the UIDs (with UGID24)!? dump/restore still don't work. Might there be a problem because I'm moving the guest systems from lvm1 to lvm2? Herbert Poetzl wrote: On Wed, Dec 21, 2005 at 10:09:21PM +0100, Lars Braeuer wrote: Hi, I'm having an issue with some files (group owner) when migrating a vserver from a 2.4.30-vs1.2.10-vquota (GID24) system, ext3 partitions mounted with tagctx to the new 2.6.14.3-vs2.0.1 system (GID24, util-vserver-0.30.209), ext3 partitions mounted with tagxid. The context id is the same (fixed) on both systems. you sure that the original partition is mounted with the proper tagxid (former tagctx) option? i.e. that the files show sane uid/gid at the 'source' When tarring the files on one vserver and untarring it on the new one, a few files have different GID's. UID's are correct as far as I can see. could you tar a few of those problematic files for me and upload or attach them to the next mail? I tarred a few files that have the wrong GID 16777214 find it at: http://mpex.net/tmp/wrong-xid.tar mtab entry (on 'source'): /dev/vg/test /vservers/test ext3 rw,tagctx,usrquota,grpquota 0 0 mtab (on 'destination'): /dev/vg/test /vservers/test ext3 rw,tagxid 0 0 on the 'source' everything looks fine. uid's/gid's are correct. here's exactly what I'm doing to transfer the files with dump. - destination: create lvm2 partition with same size as the old one - destination: create ext3 fs / mount fs with tagxid option - source: dump 0zf /tmp/test.dump /vservers/test ; copy to destination - destination: cd /vservers/test ; restore rf /test.dump - destination: Lots of GIDs are 16777214, UIDs are fine As we are not really using the tagging on the old and new system for quota (we planned to, but we never used it), is there a way to remove it before transfering the guest system via dump, i.e. with chxid -c 0 or something? Same issue when using dump/restore to transfer the guest system to the new host. The only difference is, that the GID is 16777214 on those files. just to make sure, both systems use UGID24? yep, triple checked that. Running chxid with the desired context id from the hostsystem doesn't change anything. Any hints? Am I missing anything? we'll see ... best, Herbert Regards, Lars Best, Lars ___ Vserver mailing list Vserver@list.linux-vserver.org http://list.linux-vserver.org/mailman/listinfo/vserver
[Vserver] vserver migration 1.2.10 2.0.1
Hi, I'm having an issue with some files (group owner) when migrating a vserver from a 2.4.30-vs1.2.10-vquota (GID24) system, ext3 partitions mounted with tagctx to the new 2.6.14.3-vs2.0.1 system (GID24, util-vserver-0.30.209), ext3 partitions mounted with tagxid. The context id is the same (fixed) on both systems. When tarring the files on one vserver and untarring it on the new one, a few files have different GID's. UID's are correct as far as I can see. Same issue when using dump/restore to transfer the guest system to the new host. The only difference is, that the GID is 16777214 on those files. Running chxid with the desired context id from the hostsystem doesn't change anything. Any hints? Am I missing anything? Regards, Lars ___ Vserver mailing list Vserver@list.linux-vserver.org http://list.linux-vserver.org/mailman/listinfo/vserver
Re: [Vserver] When to mount lvm? (vs2.01)
First of all, thanks for the answers. I basically understand the design of namespaces. Should have looked through this page before asking my question. Does anybody know what happens in the background if I take the example at the bottom and physically mount the device onto /vservers/test01 which will then also be used in the namespace of the vserver as mountpoint for the bind mount? Another question: There's the generateInitOptions function setting the values for INITCMD_START, INITCMD_STOP and so on. In the vserver.functions file I don't see a call for a different namespace, but test -x $vdir/vdir/etc/init.d/rc is still working (when running vserver.start). How can this be, if this all happens without a namespace set or is there somewhere a namespace call that I didn't see so far? When running vserver.stop the INITCMD_STOP can't be set, although nothing is unmounted before generateOptions in vserver.stop is called. I included a ls -la $vdir/vdir/$i right above the test -x call at line 264 (util-vserver-0.30.209) which returns: ls: /usr/local/etc/vservers/test01/vdir//etc/init.d/rc: No such file or directory Leading to a No command given; use '--help' for more information., because the INITCMD_STOP is empty. When running vserver.start it seems to see the directory: -rwxr-xr-x 1 root root 2235 Dec 13 20:11 /usr/local/etc/vservers/test01/vdir//etc/init.d/rc I know, I can set cmd.stop and everything works fine, but I want it to be detected automatically without having to set the config options myself for each guest system. Best, Lars PS: I'm running the latest stable release (2.6.14.3-vs2.01, util-vserver-0.30.209) Enrico Scholz wrote: [EMAIL PROTECTED] (Lars Braeuer) writes: I'm not sure where to mount an LVM2 device with vs2.0. I do not want to mount it on system boot, but when starting the guest vserver. The fstab file in the vserver config directory seems to be the right place. But when including an entry like this in the fstab file, the mount won't show up in the hostsystem and will not be accessible via the mount point in the /vservers directory: /dev/mapper/vg-test01 / ext3 defaults 1 2 I would recommend to do physical mounting in the host's /etc/fstab and bind-mount it in the vserver's fstab. E.g. | [/etc/fstab] | /dev/mapper/vg-test01 /srv/mnt/test01 ext3 defaults 1 2 | [/etc/vservers/.../fstab | /srv/mnt/test01/ none bind This gives the advantage that the partition will be fsck'ed during the regularly system boot and that you can share it between multiple vservers ('mount' is allowed only once per physical partition but you can 'mount --bind' multiple times). But I'm able to enter the vserver with vserver test01 enter, weird. Expected namespace behaviour... Enrico ___ Vserver mailing list Vserver@list.linux-vserver.org http://list.linux-vserver.org/mailman/listinfo/vserver
[Vserver] When to mount lvm? (vs2.0)
Hi, I'm not sure where to mount an LVM2 device with vs2.0. I do not want to mount it on system boot, but when starting the guest vserver. The fstab file in the vserver config directory seems to be the right place. But when including an entry like this in the fstab file, the mount won't show up in the hostsystem and will not be accessible via the mount point in the /vservers directory: /dev/mapper/vg-test01 / ext3 defaults 1 2 But I'm able to enter the vserver with vserver test01 enter, weird. I also tried the prepre-start script, but it seems to be the wrong place, because it's called after the generateOptions function. Any hints? Best, Lars ___ Vserver mailing list Vserver@list.linux-vserver.org http://list.linux-vserver.org/mailman/listinfo/vserver
Re: [Vserver] When to mount lvm? (vs2.0)
Chuck wrote: On Tuesday 13 December 2005 07:38 am, Lars Braeuer wrote: Hi, I'm not sure where to mount an LVM2 device with vs2.0. I do not want to mount it on system boot, but when starting the guest vserver. The fstab file in the vserver config directory seems to be the right place. good question. i run lvm2 and i have found on my system, if i want to be able to administer the guest directly from the host, i must mount it on boot or create a special script to start the guest which mounts the lvm2 mountpoint, then calls vserver to start the guest. if i do not do this, then i cannot administer the guest from the host and must enter the guest to do anything at all. will be interesting to see what the gurus say :) so you are doing it the same way, via the fstab config file? it seems like the device is mounted right until after the vserver is up. I included a ls -la /vservers/test01/ ; df -a in the prepre-start script and it actually outputs a listing of the guest system. This might be from a different ctx (i.e. 1)??? Because after the guest system is started, and I return to the normal hostsystem shell, a ls -la /vservers/test01 outputs an empty directory. But when including an entry like this in the fstab file, the mount won't show up in the hostsystem and will not be accessible via the mount point in the /vservers directory: /dev/mapper/vg-test01 / ext3 defaults 1 2 But I'm able to enter the vserver with vserver test01 enter, weird. I also tried the prepre-start script, but it seems to be the wrong place, because it's called after the generateOptions function. Any hints? Best, Lars ___ Vserver mailing list Vserver@list.linux-vserver.org http://list.linux-vserver.org/mailman/listinfo/vserver ___ Vserver mailing list Vserver@list.linux-vserver.org http://list.linux-vserver.org/mailman/listinfo/vserver
[Vserver] Docs: Migration to 2.0
hi, added a migration section to http://linux-vserver.org/Step-by-Step+Guide+2.6 let me know if you think this is helpful. it took me some time to figure out the options needed, so I thought this might be helpful for other people as well!? didn't know how to indent the code snippets so that they match the indents of the bullet points. regards, lars ___ Vserver mailing list Vserver@list.linux-vserver.org http://list.linux-vserver.org/mailman/listinfo/vserver
Re: [Vserver] Docs: Migration to 2.0
Guenther Fuchs wrote: Hi there, on Monday, December 12, 2005 at 10:46 on the list was posted: added a migration section to http://linux-vserver.org/Step-by-Step+Guide+2.6 let me know if you think this is helpful. I guess it is. IMO you should also mention to state only the nubers into the Prefix file: The Prefix (i.e. /24; check [this page] for more info) (or are both variants possible?) a thanks. that's what I wanted to do. ___ Vserver mailing list Vserver@list.linux-vserver.org http://list.linux-vserver.org/mailman/listinfo/vserver
[Vserver] vprocunhide check
Hi! Is this line in the vprocunhide init.d script supposed to check for the iattr feature? $_VSERVER_INFO - FEATURE iattr || exit 0 I remember it wasn't present in older versions. When calling vserver-info manually with these option it returns a 1 as return code. The script then stops. When commenting this line out, everything works fine. I'm using util-vserver-alpha 0.30.209. Also: One has to hide all entries by using setattr --hide --admin --~watch /proc/* (for hiding all entries within the vservers) before calling vprocunhide, right? Best, Lars ___ Vserver mailing list Vserver@list.linux-vserver.org http://list.linux-vserver.org/mailman/listinfo/vserver
Re: [Vserver] vprocunhide check
Herbert Poetzl wrote: On Thu, Dec 08, 2005 at 04:18:23PM +0100, Lars Braeuer wrote: Hi! Is this line in the vprocunhide init.d script supposed to check for the iattr feature? $_VSERVER_INFO - FEATURE iattr || exit 0 I remember it wasn't present in older versions. When calling vserver-info manually with these option it returns a 1 as return code. The script then stops. When commenting this line out, everything works fine. I'm using util-vserver-alpha 0.30.209. may I ask what kernel/patch you are using? 2.4.32 with patch for 2.4.30. I installed the alpha tools just for setattr (ProcFS security) and vprocunhide. Also: One has to hide all entries by using setattr --hide --admin --~watch /proc/* (for hiding all entries within the vservers) before calling vprocunhide, right? hmm, no, usually (i.e. the default) is to hide all entries (valid for 2.6 kernels) Ok, makes sense, cause I'm using the patch for 2.4 kernel which defaults to reveal everything. ___ Vserver mailing list Vserver@list.linux-vserver.org http://list.linux-vserver.org/mailman/listinfo/vserver
Re: [Vserver] patch for linux-2.6.14.3
Hi, I'd be interessted in a 2.6.14.3 patch as well. Tried to adapt the patch myself, but found some severe differences in several files I don't have any knowledge of, so I don't wanna mess up the stuff. ;) Best, Lars Georges Toth wrote: Hi Bertl, Could you please make a patch for linux-2.6.14.3 ? ___ Vserver mailing list Vserver@list.linux-vserver.org http://list.linux-vserver.org/mailman/listinfo/vserver
Re: [Vserver] uid problem after vserver update
Herbert Poetzl wrote: What would be the right formula for reassembling the id's for vs1.27 + quota? currently the best format would probably be UID24/GID24 #define XIDINO_UID(uid, xid) (((uid) 0xFF) | (((xid) 0xFF00) 16)) #define XIDINO_GID(gid, xid) (((gid) 0xFF) | (((xid) 0x00FF) 24)) which basically means that the upper 8 bit of (16 bit) xid go into the upper 8 bit of the (32 bit) uid, and the lower 8 bit of the xid, into the upper 8 bit of the (32 bit) gid but probably simpler would be the transition to UID32/GID16 because in this setup, you only have to mask the 'old' uid to 16 bit, and leave the gid untouched ... You mean the uid will remain untouched, since the xid goes into the upper 16 bit of the gid, correct or am I making a mistake here? Well if UID24/GID24 is the recommended format, then converting both values (uid/gid) should be as easy as converting a single one (gid), I hope. ;) Thanks for your help. Regards, Lars ___ Vserver mailing list [EMAIL PROTECTED] http://list.linux-vserver.org/mailman/listinfo/vserver
Re: [Vserver] uid problem after vserver update
Herbert Poetzl wrote: The same listing from inside the vserver (vserver name exec ls -lagn /): ... drwxr-x---2 004096 Nov 19 02:08 service drwxrwxrwt3 008192 Mar 27 19:45 tmp drwxr-xr-x 14 004096 Oct 16 11:36 usr drwxr-xr-x 20 004096 Oct 15 01:38 var When updating to the new kernel the files with the long uid's are not properly mapped anymore. We noticed this because mysql wouldn't start in any of the 10 vservers on this system. After chown'ing the related mysql.pid files and directories the mysql server started. But there are tons of files with wrong uid's left, so this wouldn't be a solution, just a temporary patch. as I see it, you have two options to migrate to 2.4.25/vs1.27 a) change the context tagging of all files to xid=0 - this can be done on the 'old' setup by using the old chctx tool, or by touching each file from the host context (xid = 0) - or with tagxid disabled, using a small script, reading the uid/gid, cropping them to 16 bits - or by doing an xid agnostic backup (tar, dump) b) converting the partition to the new context tagging - again a small script, reading the uid/gid, splitting it into xid/uid/gid and reassembling them to one of the newer formats ... So I would get the xid by dividing uid and gid by 65536, right? What would be the right formula for reassembling the id's for vs1.27 + quota? In a posting I read about three different xid tagging methods: UID32/GID16, UID24/GID24 and UID32/GID32. I couldn't find a hint when looking at the latest quota patch. ___ Vserver mailing list [EMAIL PROTECTED] http://list.linux-vserver.org/mailman/listinfo/vserver
[Vserver] uid problem after vserver update
I've had a problem when updating from 2.4.20ctx-17_quota to 2.4.24/vserver 1.26 (+quota) two weeks ago. The system hasn't been able to map the long unique uid's to usernames anymore after booting the new kernel. For now we are using the old kernel again and I'm kind of stuck in solving this problem. The first thing that seemed weird to me, is that when listing the files from outside the vserver, some files have a uid of 0 (root) and others have a long uid (i.e. 458752) which also seems to map to root (inside the vserver). Here's an example (listing via ls -lagn /vserver/name): ... drwxr-x---2 458752 458752 4096 Nov 19 02:08 service/ drwxrwxrwt3 008192 Mar 27 19:45 tmp/ drwxr-xr-x 14 004096 Oct 16 11:36 usr/ drwxr-xr-x 20 004096 Oct 15 01:38 var/ The same listing from inside the vserver (vserver name exec ls -lagn /): ... drwxr-x---2 004096 Nov 19 02:08 service drwxrwxrwt3 008192 Mar 27 19:45 tmp drwxr-xr-x 14 004096 Oct 16 11:36 usr drwxr-xr-x 20 004096 Oct 15 01:38 var When updating to the new kernel the files with the long uid's are not properly mapped anymore. We noticed this because mysql wouldn't start in any of the 10 vservers on this system. After chown'ing the related mysql.pid files and directories the mysql server started. But there are tons of files with wrong uid's left, so this wouldn't be a solution, just a temporary patch. I wasn't able to test this issue anymore, because it's on a production system. I'll be starting another try next week, so if there are any other things I could check, please tell me. Thanks in advance for your help. Regards, Lars ___ Vserver mailing list [EMAIL PROTECTED] http://list.linux-vserver.org/mailman/listinfo/vserver