Released vzctl-3.2.1CHANGES
(since 3.2)
Regressions:
* vzctl set --ram/--swap: fix (#2269)
Improvements:
* vzctl start: better error msg for vswap config
vs non-vswap kernel (#2263)
THANKS
OpenVZ project would like to thank the following people
who contributed patches and/or provided bugreports
for this vzctl release (in no particular order):
* Ilyas
* b294615@rtrtr.com
Signed-off-by: Kir Kolyshkin <kir@op...
vzctl start: better error msg for vswap config vs non-vswap kernelCommit 09c6269 added check for vswap-enabled kernel when vswap config
is to be used during vzctl start. It has been reported in bug #2263 that
some people use broken configs (that look like vswap ones) on a
non-vswap kernel, so the above commit make their CTs fail to start.
Mitigate the breakage a bit by providing more detailed info in error
message.
Should fix http://bugzilla.openvz.org/2263...
vzctl set --ram/--swap: fixCommit cd8d526 is an epic fail: if you use any suffix
the value is 1024 times larger than it should be. Oops.
Fix by dividing by _page_size. While at it, round it up to
nearest muliple of _page_size.
http://bugzilla.openvz.org/2269
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
(cherry picked from commit 136d65bc0)
Released vzctl-3.2CHANGES
(since vzctl 3.1)
New features
* vzctl console now accepts tty number argument
* vzctl console: add ESC ! to issue SAK
* vzlist: show diskspace/diskinodes usage/limit for ploop CTs
* vzlist: add more new fields
- layout (simfs/ploop)
- private/root (to show VE_PRIVATE and VE_ROOT)
- features
- smart_ctid (CT name if available, otherwise numeric CTID)
Fixes
* vzctl start: abi...
vzctl restore: do not ignore DUMPDIRAfter commit 5fb65baa vzctl restore started ignoring value
of DUMPDIR global configuration parameter. Fix it.
Reported at
http://forum.openvz.org/index.php?t=rview&goto=46169
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzmigrate: do not use --sparse for ploop CTsShould help fixing bug #2252 (although this is not a complete fix)
http://bugzilla.openvz.org/2252
Reported-by: Paparaciz <helpaz@gmail.com>
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzmigrate: simpler check for ploop CTInstead of executing vzlist we can just check the DiskDescriptor.xml
existence.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzmigrate: be silent about disabled quotaCommit f9a2ec8c made vzmigrate skip all vzquota operations for ploop
case. It was done by setting DISK_QUOTA=no which is perfectly fine,
except for the messages printed by vzmigrate saying "skipping quota
migration" (in case -v -v is specified).
We could guard those messages by enclosing them in
if [ "$PLOOP" != "yes" ]; then
...
fi
But it's easier just to remove those.
Signed-off-by: K...
vzmigrate: make -vvv add -vv to rsyncPreviously, specifying -v 3 times was used to add -v to ssh/scp.
Now with verbosity of 3 we add yet another -v to rsync, so it will show
some more information about which files to consider etc.
To have ssh/scp print debug output, specify -v four times (or -vv
two times etc.)
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzmigrate: if CT is on ploop, check it is supported on dstOtherwise we will not be able to start CT, no reason
to try migrating it.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzmigrate.in: fix check for ploop CTPrevious patch left grep spit its match to stdout which we do not want.
Also while at it let's add -w for stricter check.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzmigrate: skip vzquota ops for ploop CTsCheck if container is on ploop, and set in-script DISK_QUOTA=no,
effectively skipping all vzquota-related operations.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzlist: fix PRINT_STR_FIELDIt was broken by commit b199f80. A negative value was passed
so p_outbuffer was decremented rather than incremented.
Before this patch, vzlist could not show fields that go after
any of those for which we use PRINT_STR_FIELD.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Get rid of setup_resource_management()Just call setup_ub() directly.
Note there is no need to call check_ub() before setup_ub() because
check_ub() has already been called in vps_start_custom, no need to check
twice.
vps_run_script() is a separate beast though, let's add check_ub() to it.
This also makes the previous commit 09c6269 working, because check_ub()
was called from setup_resource_management() which was called from
vps_r...
vzctl set --ram/--swap, vzctl start: add checks for vswap-enabled kernelWhen trying to set --ram and --swap, do check that our kernel is
vswap-capable.
When trying to start container with vswap config, also check the kernel.
This simple check should prevent all sorts of confusion and frustration
I guess.
http://bugzilla.openvz.org/2251
Reported-by: Nils Breunese <nils@lemonbit.com>
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
is_image_mounted(): do not forget to free diskdescrFix a couple of minor whitespace nitpicks nearby while at it.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Introduce and use check_ploop_size()As pointed out in bug #2250, sometimes CTs come with
DISKSPACE="unlimited", which is effectlively LONG_MAX.
We are definitely not able to create a ploop device this big,
so let's add a check that it's less than some max value.
We have tried to figure out what the value is, and theoretically
it seems that it is 2 TB minus 2GB for block map and 1M for some
auxiliary data such as header. This va...
Move inc_rusage() and shift_ubs_param() to where they truly belongThese two functions are only used from inside vzmemcheck.c, so let's
- move them there (from validate.c)
- mark them as static
- remove their prototypes from validate.h
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Mark fill_2quota_param() as static, remove prototypefill_2quota_param() is only used from inside src/lib/res.c,
so it makes sense to mark it as static and remove its prototype
from res.h.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Mark get_ub_name() as static, remove prototypeget_ub_name() is only used in ub.c, so it should be marked as static.
Also let's remove its prototype from ub.h
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
include/*.h: remove non-existen function prototypesThose are probably leftovers from the removed
or changed code. Found by the following script:
for INC in include/*.h; do
for f in $(grep -E '^(int|void|char) ' $INC | awk {'print $2;'}| \
sed -e 's/(.*//' -e 's/^*//'); do
if ! git --no-pager grep -wq "$f" -- '*.c'; then
echo "$INC $f"
fi
done
done
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
build_cap_str(): do not add quotesLet's add quotes in store_cap() as for other similar functions.
Will be required for vzlist -o capabilities
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
src/lib/cap.c: fix build_cap_str() description- Fix a typo
- Remove @return since this one returns void
- Add len description
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
[style] remove NULL checks before free()free(NULL) is perfectly legal, so it doesn't make sense to check.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzctl start: remove upstart/systemd hacks for FedoraThis basically reverts commits 166d66e and 9f1963a. With the modern
OpenVZ kernels those hacks are no longer needed, plus Fedora OS
templates come with a slightly patched systemd.
NOTE that recent RHEL6-based kernel is required to run recent Fedora
distros.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
src/lib/vzfeatures.c: fix compilation with older headersWhen trying to build vzctl on a CentOS 5 system, got this (see below).
The fix is to move
#include <linux/vzcalluser.h>
to be below the
#include "util.h"
so that various defines and typedefs won't clash.
Aaaand finally, here's the error message:
CC vzfeatures.lo
In file included from /usr/include/stdlib.h:438,
from ../../include/util.h:21,
from vz...
Move struct log_param definition from logger.h to logger.c- Now we do not need to include <stdio.h> to logger.h
- Next thing we see, we do not have this include in enter.c and ub.c
- So add it there
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Move osrelease-related functions to where they belongCommit f0a858dd introduced support for setting osrelease.
Functions used by it are only used in env.c, so let's move them
there and make them all static.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzlist: do not use /proc/bc/resources for non-running CTsFor stopped CTs, UBC parameters should NOT be obtained from
/proc/bc/resources. Since /proc/bc/resources do not usually
list stopped CTs, this usually never happens.
In some rare cases though a container is stopped but for some
of its beancounters held value is not zero (such as because
of a kernel bug in UBC accounting), and its UBC entry remains.
In such a case, values from the proc file sho...
vzlist: introduce and use FOR_ALL_UBC macroThis is mostly to save a few lines of code and make sure that
if we ever add a new beancounter it will be easier to add it
to vzlist code.
Unfortunately we can't use this macro with UBC_FIELD.
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzlist: refactor get_ub()/update_ubc()1. Do not use struct Cveinfo -- there is no need to
2. Only allocate new struct Cubc when it's really needed
Number 2 should fix a lot of memory leaks :)
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
vzlist: show diskspace/diskinodes usage/limit for ploop CTsWe are using information from /proc/vz/vzquota to show
diskspace and diskinodes usage and limits for CTs. Of course,
it is not working for ploop-based CTs.
So let's use ploop_info() for containers on ploop. Note that,
unlike vzquota, it works for stopped CTs as well.
BEFORE THIS PATCH:
# vzlist -o ctid,layout,status,diskspace,diskspace.h,diskinodes,diskinodes.h -a
CTID LAYOUT STATUS ...
vzlist: fix showing disk usage for mounted CTsIn case a container is mounted (and quota is enabled for it),
its diskspace/diskinodes usage is well known from /proc/vz/vzquota.
But because of status == VE_RUNNING check in PRINT_DQ the usage
is now shown.
Fix it by replacing the "is running" checking to "is usage==0".
Signed-off-by: Kir Kolyshkin <kir@openvz.org>
Kir Kolyshkin