Released vzctl 3.0.24.1This is a stable update to 3.0.24, fixing some regressions:
* vzctl start --wait: fix for Fedora 13
* fix running default distscripts if OSTEMPLATE is not set
* vzmigrate: migrate premount/postumount action scripts
* vzctl set --hostname: back to FQDN for Red Hat and Slackware
* vzmigrate: ignore error 24 from the first rsync run
* bash_completion: vzlist's stderr goes to /dev/null
Signed-off-...
bash_completion: vzlist's stderr goes to /dev/nullvzlist can write some warnings to stderr. This is not
what we want to see while working with bash_completion.
Example of what could happen before applying this patch:
vzctl create <tab>Invalid value for IOPRIO=88, skipped
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzmigrate: ignore error 24 from the first rsync runNow, when we really started to check the exit code from rsync, we got
the following problem. Apparently, sometimes first rsync run fails
with error 24, it happens when a file to be transferred is removed.
Since the container is running at this time there is nothing wrong
with it, so ignore error 24.
Should fix bug #1580.
http://bugzilla.openvz.org/1580
Reported-by: Roman Veretelnikov <roman@...
vzctl set --hostname: back to FQDN for Red Hat and SlackwareCommit 57bfaa5a61064 was a bit radical in that it removed the
FQDN from all of the places.
As many people pointed out, Red Hat's /etc/sysconfig/network
is expected to have HOSTNAME= value to be FQDN, for example,
see [1]. Fix that by reverting a hunk in the above commit.
While at it, I have checked if the issue is valid for other distros:
1. Gentoo: /etc/conf.d/hostname should be hostname-on...
vzmigrate: migrate premount/postumount action scriptsCommit eeceb721472483f6e8d90cbfe72967c3336b5e77 added
support for premount/postumount action scripts. vzmigrate
should know about these scripts in order to migrate them.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
(cherry picked from commit b8aa99670ac2905b3a4ed44edd75d259e7b258be)
Revert "Check for memory granted by get_dist_name"This reverts commit 80ad89e384e04bd0835a13f246840f4f71113263.
The commit have a few problems, unfortunately I overlooked it.
1. return NULL from get_dist_name() is perfectly valid,
it means that both tmpl->dist nor tmpl->ostmpl are NULL.
This NULL is later handled by read_dist_actions().
With this patch return NULL from get_dist_name() becomes
an error, but there is no error!
2. ...
vzctl start --wait: fixed order of checking for upstart config dirsWe need to check for /etc/init/ (new style upstart) before /etc/event.d/.
This should fix --wait on Fedora 13.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
(cherry picked from commit ac4a19ff2c6de9a2ea56eeeeaca14ea738ed5b2d)
Released vzctl 3.0.24CHANGES
* Added new swappages UBC parameter
* Added new vzctl set --features: ppp, ipgre, bridge, nfsd
* Feature 'sysfs' is now on by default
* Added new CT parameter: bootorder (#1300)
* Added delete alias for destroy command
* Added unlimited CT config (#1080)
* Added support for vps.create global user script
* Added support for premount and postumount global and per-CT user scripts
* Added ...
vzmigrate: saner output for non-existing CTIDCommit 484afba08 added support for CT names. It works, but
one corner case is not working particularly well. When non-existing
numeric CT ID is supplied, vzlist prints "Container(s) not found"
error which is a bit misleading. (If non-numeric non-existing ID
is supplied to vzlist the error is "CT ID 12434a is invalid"
which is fine in the context of running vzmigrate.)
To fix this problem, dd a...
vzmigrate: saner output in case of absent configThere is a statement that checks the readability of both
global and CT configs in one sitting. In case of problem
it reports:
Can't read global config or CT #$VEID config file
This message is a bit strange, because it's not clear what file is
missing.
To fix, remove the check for per-CT config file and fix the error
message accordingly. Check for per-CT config is done later, when
vzctl st...
vzmigrate: assorted minor cleanups(1) Fix a typo in a comment.
(2) Do not log option values, it makes no sense.
(3) Do not brake the line shorter than 80 characters.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzmigrate: fix loggingCommit ea203fb5bc changed echo to printf because echo -e is not
portable. Unfortunately it did it in a bad way:
(1) no newline character
(2) bad number of arguments to printf
Revert it and go back to standard echo. But do not use -e now,
fixing a few statements than relied on it by changing \t to spaces.
Use $* and put it inside the doubple quotas to save these spaces.
Signed-off-by: Kir Ko...
vzctl set --ipdel all: fix removing IPs from CTThis is a bit tricky. vps_net_ctl() is called twice -- for ADD
and for DEL operations, with different resources ('net' and 'del_net').
The problem is delall flag is only set in 'net' but not in 'del_net'.
When vps_net_ctl() is called for DEL it checks that the list of IPs
is empty and 'delall' is not set (because it is not set in del_net)
so it exits immediately, therefore vps_ip_configure(op=...
vzlist: clean up some unneeded RES_ definesDefines RES_NAME, RES_DESCRIPTION, RES_ONBOOT, RES_BOOTORDER and RES_IO
are completely unnecessary. Remove them.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
src/vzlist.c: streamline a few macrosSimplify macro calls by shortening the list of arguments, add some
macros to avoid repetition, and rename a few disk quota releated
functions.
Presumably this does not change the generated code in any way.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzlist: fix headers for disk quota fieldsThe appropriate parameters are diskspace and diskinodes, therefore
let's name the fields as DSPACE and DINODES, not DQBLOCKS and DQINODES.
Fix the man page appropriately.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzlist: put similar print_ functions in a macroFunctions print_name(), print_hostname() and print_description()
are all monozygotic twins. Create a macro to generate those.
While at it, move these functions plus print_ip up in the source
file so we no longer need prototypes.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
man/vzlist.8: fix rendering for ioprio in a tableFix to previous commit 53b271ab1f8880504c3b81297e3b5b778653e737.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzlist: add ioprio fieldCurrently the value is only obtained from CT config files.
We could add support for getting the value from /proc/bc/$CTID/ioprio
some time later.
http://bugzilla.openvz.org/show_bug.cgi?id=1526
Reported-by: Dennis P. Nikolaenko <dennis@nikolaenko.ru>
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
parse_ioprio(): reset ioprio if value is invalidIf ioprio value is not within allowed range, do not just report
an error, also reset the value to -1 (which means 'unset').
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
etc/dists/scripts/*: Get rid of 'echo -e' bashismsApparently using 'echo -e' is non-portable. In these scripts
this is harmless since they are to be run by bash (first line
specifically says #!/bin/bash). But since -e functionality
(i.e. interpretation of backslash sequences line \n and \t)
is not required here, just remove '-e' flag.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
scripts/vpsnetclean: Get rid of 'echo -e' bashismApparently using 'echo -e' is non-portable. More to say,
in these two cases -e is not needed, so just remove '-e'.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
scripts/vpsreboot: use vzlist for list of stopped CTsGet rid of fancy trickery, just use vzlist -H -octid -S
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzmigrate: Get rid of 'echo -e' bashismApparently using 'echo -e' is non-portable. Replace to printf
which is supposed to be portable and understand escape sequences
(as recommended by https://wiki.ubuntu.com/DashAsBinSh).
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=581147
Reported-by: Raphael Geissert <atomo64@gmail.com>
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzctl destroy: emit error if failed to rmdir VE_ROOTThis is a fix/addition to the previous commit 560f0a33.
It's nice to print errno as well.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzctl destroy: emit warning if unable to remove VE_ROOTVE_ROOT is normally an empty directory, so we call rmdir() on it.
But in case it is not empty, or not a directory, rmdir fails.
This behavior is totally correct, but it's nice to let the user know
about the problem.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzctl destroy: handle the case when VE_PRIVATE is a symlinkIn case VE_PRIVATE is a symlink let's print its destination
and only try to remove the symlink itself, not its destination.
Not removing symlink destination is the common-sense policy
described in symlink(7) and obeyed by utilities such as rm(3).
http://bugzilla.openvz.org/1095
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Build system: generate files silentlyI.e. use AM_V_GEN for autogenerated files. In case current automake
does not support silent build, it will expand to nothing so no harm
in any case.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
init.d/vz-redhat: warn user if IP forwarding is offThis was already implemented for Gentoo but not for Red Hat/SUSE/Debian.
http://bugzilla.openvz.org/1079
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
init.d/vz-gentoo: simplify check for IP forwardingCommit 2b97841cd25a8088d5b878eeb5420e5509dc142e added checking
if IP forwarding is enabled. Use sysctl options to avoid using awk
and /dev/null redirection, also explicitly check for value of 1.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
init.d/vz-redhat: make it less verboseJust print that we start venet0, no need to report every
ip and sysctl action.
Also, removed misleading FIXME comment -- address assigned
is the correct one.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzctl destroy: fix rare off-by-one error in get_destroy_root()In case VE_PRIVATE is a separate file system (say you mount
/dev/sda3 to /vz/private/333), or VE_PRIVATE is a symlink to
a directory on another filesystem (like
/vz/private/333 -> /otherdir/333), function get_destroy_root()
should return the same path it is given. Now it cuts the last
character as shown below:
$ ls -l /vz/private/801
lrwxrwxrwx 1 root root 13 May 12 20:43 /vz/private/801 -> /o...
vzctl destroy: fix a locking bugIf the lock is already taken we do not need to unlock it,
just silently exit (someone else is already doing our job).
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzmigrate: do not try to turn quota off if DISK_QUOTA=offThe bug prevents migrating stopped CT when DISK_QUOTA=off.
We check DISK_QUOTA in other places but forgot it here.
Initially reported by Valerij Frolov.
Initial fix by Marat Stanichenko.
http://bugzilla.openvz.org/1094
vzctl start: mount /proc and /sys inside CTSome of the modern distros rely on initrd scripts doing this,
and since CTs don't have initrd we do it in vzctl.
Another reason is commit 98642b9e7a640a9c989898b69870effcbbfb651e
which tries to mount to /proc/fs/nfsd so we need to have /proc
mounted before.
Also, move mk_quota_link to after mounting /proc and such,
because it involves /proc (as a link destination so it doesn't
really matter b...
vzctl set --features nfsd: recognize itSince feature "nfs" is before "nfsd" in the array of features,
we were getting false positive in comparison, then following
check for comma was negative and the result is "unknown feature".
Fix that by moving the check for comma next to strncmp.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Mount nfsd related stuff on CT start if nfsd feature is enabledThese two:
nfsd at /proc/fs/nfsd
rpc_pipefs at /var/lib/nfs/rpc_pipefs
should be mounted at container boot in order to make nfs startup
scripts work.
bin/.gitignore: add vznetaddbrCommit 41fb6973bc205c00e25ba73431110ac8e821d6c9 adds vznetaddbr
script (autogenerated from vznetaddbr.in), but forgets to add
vznetaddbr to .gitignore. Do it.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Enable sysfs feature by defaultMost of the newer distros need it, so enable it. User can still
disable it explicitly if needed, by using --features sysfs:off.
The feature was not on by default because older kernels used to
pin about 2M of kernel memory per sysfs instance for container.
Now it's not the case so we can enable it harmlessly.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Use enum instead of defines for ACTION_sThis makes compiler do some stricter checking,
also debugging is easier since you see the name not value.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Kir Kolyshkin
