Commits
Author | Commit | Message | Commit date | Issues | |
---|---|---|---|---|---|
Kir Kolyshkin | c90c3dc3bae | Released vzctl 4.1Changes: see http://wiki.openvz.org/Download/vzctl/4.1/changes Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | 0e6293370cf | setver.sh: if rpm install failed, exit with error... and do not remove rpm files. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | 4794d7e726b | setver.sh: add -o|--oldpackage optionSigned-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | db7900e08d4 | vzevent-stop: workaround for F17 that doesn't call reboot()The problem is systemd from Fedora 17 doesn't invoke reboot() syscall at all, so kernel doesn't know if CT is rebooted and send us stop event instead of reboot event. The workaround has two parts 1 vzdummy-systemd-f17 package installed inside an F17-based container have /usr/lib/systemd/system/vzreboot.service file (and a symlink to it from /lib/systemd/system/reboot.target.wants/) which ... | |||
Kir Kolyshkin | b8a4d959628 | suse-add_ip.sh: only set default route to venet0 if there is no otherAs per the bug report, only set default route to venet0 if there is no any other default route set. http://bugzilla.openvz.org/2376 Reported-by: Andres Toomsalu <andres@active.ee> Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | 8c287191ce2 | etc/init.d/vz*: eliminate "Container(s) not found" msgWhen doing vz stop, most probably all CTs will suspend fine during the first stage, so after it vzlist return nothing but do this: ++ vzlist -H -o ctid -sbootorder Container(s) not found To fix, redirect its stderr to /dev/null Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | a1dd7167e38 | setver.sh: do not keep tmp files behindDo move not copy when restoring original configure.ac and vzctl.spec Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | af5bf4303c5 | etc/init.d/vz*: restore running containers after rebootOn vz stop, save CTIDs of all running containers to /var/lib/vzctl/vzreboot On vz start, start containers with saved IDs (plus the ones marked with ONBOOT=yes) Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | 8fa8a6193c3 | etc/init.d/vz* stop: add pre-stop stage(1) Before stopping/suspending all CTs, set the same value for their cpuunits for them to have equal abilities to stop. Before this patch, we were setting cpuunits for CTs one by one, and high cpuunits value for some CTs that stop later might cause a slowdown. (2) Unset cpulimit only before stopping CT (doesn't make sense to do before suspending). Signed-off-by: Kir Kolyshkin ... | |||
Kir Kolyshkin | 8cf2db4a2ad | etc/init.d/vz*: fix vzlist invocation in stop_ve(s)Option -H was forgotten. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | 6812ff71787 | properly propagate /var/lib/vzctl/veip dirOtherwise we could run into a discrepancy then Makefile thinks it's /usr/var and vzctl thinks it's /var. I am not sure why it should be /usr/var and whoever on earth uses this, but it's the same problem as with /etc vs /usr/etc (which is workarounded in ./configure script). While at it, also remove non-relevant comment about iptables from scripts/vps-functions.in Signed-off-by: Kir Kolyshkin... | |||
Kir Kolyshkin | 0aef01baf4c | etc/init.d/vz*: fix printing literal $veidIt was like this: # /etc/init.d/vz stop Suspending CT $veid Suspending CT $veid Suspending CT $veid Suspending CT $veid Suspending CT $veid Suspending CT $veid Suspending CT $veid Suspending CT $veid Fixup after commit 537888a. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | f14a6e11980 | etc/init.d/vz-redhat: add more locals to stop_vesFixup after commit 537888a. For some reason the gentoo script was OK. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | b5130d5a094 | vps-create: fix checking needed disk spaceOh this is quite a story. Sit down and listen, folks. First, since 14 Dec 2002 (coreutils-4.5.4) and up until to 6 Jan 2012 (coreutils-8.15), ls -lk was printing file size in kilobytes. It's man page (as of coreutils-8.15) was still saying that: -k, --kibibytes use 1024-byte blocks although info page is already fixed: `-k' `--kibibytes' Set the default block size... | |||
Kir Kolyshkin | c3cb0b21120 | vzctl restore: print 'container restore failed'If fn is not NULL in vps_start_custom(), it means we're doing restore rather than start. Fix log messages accordingly. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | e7b64797a88 | vzctl restore: do not print "Starting container"Move "Starting container..." message from vps_start_custom() to vps_start(),so when doing restore we print "Restoring", and when doing start we print "Starting". Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | f95b00bbbed | setver.sh: remove dist tarball and (s)rpmsIf we do install we probably don't need dist tarball and just-installed RPMS. Do clean those. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | 70fc6606775 | setver.sh: do not hardcode vzctl nameSigned-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | 0efd039fcc8 | Makefile.am: we only need .bz2 for make rpmsWe use tar.bz2 as a source archive to make rpms, so depend on dist-bzip2 rather than dist (which makes both tar.gz and tar.bz2). Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | 0f02c03f908 | vzctl start: do not remount fs if we try restore firstWhen vzctl start tries to restore a CT, it mounts it first, then, if restore fails, it unmounts and then tries to start and mounts again. Introduce SKIP_UMOUNT and SKIP_REMOUNT flags and use them to skip unecessary remount. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | aa23dde76a7 | start(): refactor setting skip flagNeeded for the next patch Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | a90feba57eb | vzctl start: move checks from try_restore() to start()Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | e7ae4d4a094 | vzctl start: tell we're going to restorePrint "trying to restore" and the dump file name. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | ad54db6d215 | vzctl start: do start if restore failedOtherwise we are leaving CT in a non-startable state. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | e916962a0cc | setver.sh: remove -c|--clean option, set it automaticallyWe need clean when we do build. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | 9e3413f20c7 | vzctl_resize_image(): initialize ploop_resize_paramSigned-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | cab43d9ae39 | vzeventd: do process -h optionSomehow it was omitted from getopt() argument. Help was still printed, the only consequences were: 1 error message saying -h is unknown option 2 exit code of 1 instead of 0 Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | c31afc7e2b9 | setver.sh: add -c|--cleanTo restore vzctl.spec and configure.ac to originals. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | 15e8b70bfcf | vzctl snapshot-switch: fix ve.conf restorationYet another stupid mistake. I swear I was not on drugs! Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | 8f9599793d5 | vzctl snapshot*: improve --id parameter parsing1. Allow for GUID not in curly brackets {} 2. Validate GUID format. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | 0519dfbed0f | vzctl snapshot: fix copying CT configIt was a very stupid typo of mine that resulted in copying Snapshots.xml instead of CT config file when creating a snapshot. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | bbb68633df6 | set_userpass.sh: fix bashismUse sed to split user:passwd string. Originally reported to Debian: > possible bashism in ./etc/vz/dists/scripts/set_userpass.sh line 26 > (${parm/?/pat[/str]}): > local user=${userpw/:*/} http://bugzilla.openvz.org/2403 http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=690713 Reported-by: Raphael Geissert <atomo64@gmail.com> Reported-by: Ola Lundqvist <opal@debian.org> Signed-off-by: ... | |||
Kir Kolyshkin | 845a257aaaf | scripts/vps-download: fix bogus warning from checkbashismsIt complains heredocument construct is not closed. Changing -EOF- to =EOF= helps ;) Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | d200fc1e89f | Add OOM adjustmentsSince kernel 042stab056 we have a kernel mechanism to auto-adjust process oom_score based on it's name, parent name, and pid. This is very useful in multi-container system, especially for old distros which are not oom-score aware. OOM groups configuration is in /etc/vz/oom-groups.conf which is applied to the kernel by /etc/init.d/vz script on start. References: PSBM-12263, PSBM-13729 Signed-o... | 2 Jira Issues | ||
Kir Kolyshkin | e7d97c5997c | etc/Makefile.am: don't list conf files twice, add CLEANFILESSince commit 3dd31a1 vz.conf is autogenerated from vz.conf.in, while other .conf files are not. Use make feature to remove the .in suffix in order to not list the files twice. Also, add vz.conf to CLEANFILES since it's autogenerated. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | 537888a00cc | etc/init.d/vz*: implement suspend instead of stopFirst, we try to suspend CTs, then we stop ones that were failed to suspend. Note that a recent patch adds restore capability to 'vzctl start', so we don't need to change start_ve() http://bugzilla.openvz.org/2325 Reference: PSBM-12516 Signed-off-by: Kir Kolyshkin <kir@openvz.org> | PSBM-12516 | ||
Kir Kolyshkin | c0bf9cd73c2 | etc/init.d/vz-redhat: add missing locals to stop_ves()And put them all to a single line Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | b70316497f5 | etc/init.d/vz*: simplify stop_veRemove the outer if statement by returning early. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | a4f17001e3d | /etc/init.d/vz-*: separate umount from stop of CTsThis way it is easier to read and modify. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | 7862fc4534a | getlockpid(): fix potential buffer overflowread() can read and return up to sizeof(buf) bytes, and then buf[sizeof(buf)] is out of bounds. Can't happen in reality, but who knows. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | dd378c0f9de | If a CT is locked, show pid and cmdline of a lockerReference: vzctl@807236 Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | d56fd7d7483 | vzctl umount: fix for case when CT have deleted mount pointsIf a container have deleted mount points inside, they are shown with '(deleted) ' prefix and thus can't be unmounted by umount_submounts(). Fix by removing the prefix. Reference: vzctl@807307 Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | 2efbc36db28 | vzctl start: try to restore container firstIf a container is suspended (default dump file exist), vzctl start tries to restore it first. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | f8fb3745a0b | vzctl set --mount_opts: fix a segfaultThis (pretty stupid) bug was introduced in commit 287d0dc, and led to segfault when initially setting mount options for a running container. http://bugzilla.openvz.org/2385 Reported-by: Dmitry Kopytov <kopytov@kopytov.ru> Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | 29a73504a17 | vzctl(8): improved vzctl create --layout/--diskspace descriptionFor --layout ploop, state that using anything else but default is neither recommended not supported. For --diskspace, state that it is required for ploop in case DISKSPACE is not set in configs. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | 11b3e64bad4 | vzctl(8): improve --diskspace description.Remove the notion of blocks, it's just easier to say "kilobytes". Mention that for ploop the second value is ignored. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | 3aa97496030 | vzctl: add 'suspend' and 'resume' aliases- suspend is the same as chkpnt - resume is the same as restore bash-completion script and man page modified accordingly. NOTE we are not patching vzmigrate script, since it calls vzctl on the other side and there could be an older vzctl version that is not aware of these new aliases. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | dc96c59faf7 | vzctl.spec: supply vzdir to configureThe value of %_vzdir set in vzctl.spec is now passed on to ./configure script, meaning one can change it and still have a proper built rpm. Signed-off-by: Kir Kolyshkin <kir@openvz.org> | |||
Kir Kolyshkin | 50b988d7090 | configure: add ability to alter /vz pathWith the help of a few previous patches, we can now change /vz while doing ./configure. Unfortunately, custom options available to ./configure are limited to --with-something and --enable-something. Neither works for this case, so we are using environment variable: ./configure vzdir=/some/dir and (with the help of AC_ARG_VAR) vzdir will end up in Makefiles. This fixes bug #421! http://bug... | |||
Kir Kolyshkin | f55bdf676e0 | include/types.h: do not hardcode /vz/dumpUse VZDIR from paths.am Signed-off-by: Kir Kolyshkin <kir@openvz.org> |