Thierry Parmentelat [Thu, 12 Nov 2015 19:01:41 +0000 (20:01 +0100)]
turn off the codemux plugin in an lxc environment
it was confirmed to be responsible for the slice re-creation issue
Thierry Parmentelat [Thu, 12 Nov 2015 19:01:03 +0000 (20:01 +0100)]
cosmetic changes only, pretty printing and comments, no change in behaviour
Thierry Parmentelat [Tue, 10 Nov 2015 15:04:37 +0000 (16:04 +0100)]
trace actions done on cgroups so we can check if this happens within the lifespan of the sliver
Thierry Parmentelat [Wed, 4 Nov 2015 11:09:56 +0000 (12:09 +0100)]
cosmetic - slightly changed message when a freezer cgroup file can't be figured out
Thierry Parmentelat [Wed, 4 Nov 2015 10:54:26 +0000 (11:54 +0100)]
remove noise in nodemanager log
Thierry Parmentelat [Wed, 4 Nov 2015 10:39:44 +0000 (11:39 +0100)]
reconnect to libvirt when the (unique) connection object looks broken
Thierry Parmentelat [Mon, 2 Nov 2015 15:09:28 +0000 (16:09 +0100)]
se format() instead of % in sliver_lxc.py
Thierry Parmentelat [Mon, 2 Nov 2015 13:48:32 +0000 (14:48 +0100)]
logging : when a reference image is not found, its expected path was not printed out properly
Thierry Parmentelat [Fri, 26 Jun 2015 12:00:35 +0000 (14:00 +0200)]
Setting tag nodemanager-5.2-17
first step towards auto slice-reimaging
slices that would be reimaged because of a change of slicefamily are
identified in the nodemanager's log
might have fixed fd leaks (use context managers whenever possible)
Thierry Parmentelat [Fri, 19 Jun 2015 15:27:28 +0000 (17:27 +0200)]
reimaging code revisited
Thierry Parmentelat [Fri, 19 Jun 2015 15:10:16 +0000 (17:10 +0200)]
spaces after comma
Thierry Parmentelat [Fri, 19 Jun 2015 14:45:03 +0000 (16:45 +0200)]
first stab at reimaging slivers when created under the wrong slicefamily
Thierry Parmentelat [Fri, 19 Jun 2015 12:20:21 +0000 (14:20 +0200)]
hopefully fewer leaks in files : always use a context manager to open files
Thierry Parmentelat [Fri, 19 Jun 2015 10:54:28 +0000 (12:54 +0200)]
another pass for prettifying - mostly format instead of %
Thierry Parmentelat [Fri, 19 Jun 2015 09:53:32 +0000 (11:53 +0200)]
no change, just nicer
Thierry Parmentelat [Fri, 3 Apr 2015 07:55:24 +0000 (09:55 +0200)]
Setting tag nodemanager-5.2-16
initscript plugin reviewed
turn off the initscript machinery completely on hosts that run f>=20
some tweaks related to cgroup naming in f>=20, including for ipv6
ideally all code depending on the cgroup naming scheme should use cgroups.py
we are not there yet but this is a bit cleaner already
Thierry Parmentelat [Fri, 3 Apr 2015 06:32:32 +0000 (08:32 +0200)]
cosmetic tweaks before tagging
Guilherme Sperb Machado [Wed, 1 Apr 2015 00:40:16 +0000 (02:40 +0200)]
refactor of reboot_slivers() to cover f21 -- only ipv6 part is affected.
Guilherme Sperb Machado [Mon, 23 Mar 2015 22:17:09 +0000 (23:17 +0100)]
Correted a bug in the get_sliver_process() function in tools.py to
consider libvirt cgroup naming standards in Fedora 20 and later. Clean
up a bit ipv6.py and update_ipv6addr_slivertag.py, also correcting minor
bugs.
Thierry Parmentelat [Fri, 20 Mar 2015 09:48:42 +0000 (10:48 +0100)]
replace confusing warning message with a debug one
Thierry Parmentelat [Wed, 18 Mar 2015 20:07:10 +0000 (21:07 +0100)]
turn off the initscript feature altogether in nodes that have machinectl (i.e. f20 and higher)
Thierry Parmentelat [Wed, 18 Mar 2015 10:55:05 +0000 (11:55 +0100)]
an attempt to see if avoiding to run initscripts altogether solves the slice re-creation issue
Thierry Parmentelat [Fri, 13 Mar 2015 15:35:09 +0000 (16:35 +0100)]
cleaned up cgroups, updated for recent libvirts,
used in coresched - at least the part that keeps on sending messages;
don't know how to check the reservation part
Thierry Parmentelat [Fri, 13 Mar 2015 14:23:45 +0000 (15:23 +0100)]
pretty-printed
Thierry Parmentelat [Fri, 13 Mar 2015 11:02:24 +0000 (12:02 +0100)]
deprecate refresh_slice_vinit and move this code up to Initscript.configure
Thierry Parmentelat [Fri, 13 Mar 2015 11:01:19 +0000 (12:01 +0100)]
minor cleanup
Thierry Parmentelat [Fri, 13 Mar 2015 10:59:54 +0000 (11:59 +0100)]
fix coresched locating cgroup and reduce verbosity
Thierry Parmentelat [Fri, 13 Mar 2015 10:29:13 +0000 (11:29 +0100)]
oops, too much messing about
Thierry Parmentelat [Fri, 13 Mar 2015 09:53:21 +0000 (10:53 +0100)]
trim down the amount of verbosity of update_ipv6addr_slivertag
Thierry Parmentelat [Fri, 13 Mar 2015 09:52:49 +0000 (10:52 +0100)]
let SystemExit exception go up to the surface so we can exit() and so freeze the system for troubleshooting
Thierry Parmentelat [Fri, 13 Mar 2015 09:47:58 +0000 (10:47 +0100)]
just cosmetics
Thierry Parmentelat [Fri, 13 Mar 2015 07:26:23 +0000 (08:26 +0100)]
oops
Thierry Parmentelat [Thu, 12 Mar 2015 20:23:47 +0000 (21:23 +0100)]
minor tweaks, and revert on forcing refresh_slice_vinit, this is not the problem
Thierry Parmentelat [Thu, 12 Mar 2015 15:46:59 +0000 (16:46 +0100)]
a few fixes about initscripts under systemd
Thierry Parmentelat [Thu, 12 Mar 2015 15:38:10 +0000 (16:38 +0100)]
more pythonic spaces - no real change
Thierry Parmentelat [Thu, 12 Mar 2015 09:39:38 +0000 (10:39 +0100)]
systemctl --system daemon-reload
Thierry Parmentelat [Tue, 10 Mar 2015 14:42:12 +0000 (15:42 +0100)]
another attempt for smoother initscript integration
Thierry Parmentelat [Tue, 10 Mar 2015 14:41:37 +0000 (15:41 +0100)]
a convenience target to push systemd unit files on the sliver side
Thierry Parmentelat [Tue, 10 Mar 2015 14:40:59 +0000 (15:40 +0100)]
only cosmetic
I have noticed numerous failures in get_sliver_process though
Thierry Parmentelat [Wed, 18 Feb 2015 10:31:46 +0000 (11:31 +0100)]
Setting tag nodemanager-5.2-15
addition to support for ipv6 to lxc slivers
(thanks to Guilherme Sperb Machado)
privatebridge plugin is back to the -lxc rpm
add systemd dependency to network.target
Thomas Dreibholz [Fri, 6 Feb 2015 19:53:08 +0000 (20:53 +0100)]
Added debug code.
Thomas Dreibholz [Fri, 6 Feb 2015 15:14:44 +0000 (16:14 +0100)]
Added lsof call for debugging the F20 issue.
Thierry Parmentelat [Thu, 30 Oct 2014 08:58:39 +0000 (09:58 +0100)]
Merge branch 'ipv6'
only one minor conflict found in a comment in Makefile
Guilherme Sperb Machado [Fri, 3 Oct 2014 13:06:06 +0000 (15:06 +0200)]
small updates here and there in all three files. The
update_ipv6addr-slivertag.py and ipv6.py were updated to support VS,
since "import libvirt" raises error and fails the build on -f14.
Thierry Parmentelat [Fri, 3 Oct 2014 10:36:04 +0000 (12:36 +0200)]
new syncvs target
Guilherme Sperb Machado [Mon, 29 Sep 2014 13:37:24 +0000 (15:37 +0200)]
ipv6.py adapted to check if the node's virtualization is LXC or VS,
update_ipv6_slivertag.py adapted to not always remove ipv6 addrs on
/etc/hosts, and tools.py adapted related to import's and search of ipv6
addrs.
Thierry Parmentelat [Thu, 25 Sep 2014 18:59:22 +0000 (20:59 +0200)]
move the 2 ipv6-related plugins over to the common packaging
Thierry Parmentelat [Tue, 23 Sep 2014 15:35:12 +0000 (17:35 +0200)]
an attempt to put the privatebridge plugin back into the -lxc rpm
Thierry Parmentelat [Tue, 23 Sep 2014 15:32:52 +0000 (17:32 +0200)]
tweak privatebridge to remove exception message when ovs is not installed - prints out warning instead
Thierry Parmentelat [Tue, 23 Sep 2014 15:35:12 +0000 (17:35 +0200)]
an attempt to put the privatebridge plugin back into the -lxc rpm
Thierry Parmentelat [Tue, 23 Sep 2014 15:32:52 +0000 (17:32 +0200)]
tweak privatebridge to remove exception message when ovs is not installed - prints out warning instead
Guilherme Sperb Machado [Fri, 19 Sep 2014 12:53:27 +0000 (14:53 +0200)]
Bug fix on the update_ipv6addr_slivertag.py plugin. Just avoided to use
GetSliceIPv6Address() method.
Guilherme Sperb Machado [Thu, 18 Sep 2014 16:34:46 +0000 (18:34 +0200)]
few changes on the ipv6.py logic, and a bit of code cleanup in tools.py,
update_ipv6addr_slivertag.py, and sliver_lxc.py.
Guilherme Sperb Machado [Thu, 18 Sep 2014 00:18:01 +0000 (02:18 +0200)]
modified some logic in the ipv6.py (bug fixing) and naming fix in the
tools.py
Guilherme Sperb Machado [Tue, 16 Sep 2014 23:38:50 +0000 (01:38 +0200)]
major indent mistake on a for loop
Thierry Parmentelat [Tue, 16 Sep 2014 09:48:14 +0000 (11:48 +0200)]
reviewed Guilherme's code
had to deeply modify indentations as these were coming with tabs and looked like nothing within my emacs
fixed the issue in tools.py on f14/vs : logger was used before imported
Thierry Parmentelat [Mon, 15 Sep 2014 06:06:28 +0000 (08:06 +0200)]
iron out the sync target - not used at build time
Thierry Parmentelat [Mon, 15 Sep 2014 06:06:10 +0000 (08:06 +0200)]
add ipv6 and update_ipv6addr_slivertag to the lxc rpm
Guilherme Sperb Machado [Sun, 14 Sep 2014 14:08:26 +0000 (16:08 +0200)]
modified ipv6.py and update_ipv6addr_slivertag.py -- identation, removed unused methods/variables
Guilherme Sperb Machado [Sat, 13 Sep 2014 11:11:32 +0000 (13:11 +0200)]
just an identation error correction
Guilherme Sperb Machado [Fri, 12 Sep 2014 14:47:59 +0000 (16:47 +0200)]
just modified the tools.py
Guilherme Sperb Machado [Fri, 12 Sep 2014 14:38:06 +0000 (16:38 +0200)]
added files related to the ipv6 support on slices
Thomas Dreibholz [Tue, 22 Jul 2014 06:22:42 +0000 (08:22 +0200)]
nm.service fix: the node manager needs network.target, since it needs to contact the PLC.
Thierry Parmentelat [Wed, 16 Jul 2014 10:20:44 +0000 (12:20 +0200)]
Setting tag nodemanager-5.2-14
review packaging : one single spec file
review packaging : some cleanup towards debians although not complete yet
minor tweaks in exposing user's .ssh dir readonly
Thierry Parmentelat [Wed, 9 Jul 2014 12:37:48 +0000 (14:37 +0200)]
just added a comment
Thierry Parmentelat [Wed, 2 Jul 2014 08:05:12 +0000 (10:05 +0200)]
fix the way 'arch' is computed, as exposed to the guest XML template
Thierry Parmentelat [Wed, 2 Jul 2014 07:56:50 +0000 (09:56 +0200)]
no functional change, just comments + cosmetic code change to be more similar to the lxc code
Thierry Parmentelat [Mon, 30 Jun 2014 14:22:53 +0000 (16:22 +0200)]
Thomas's change to ensure tighter synchro between conf_files lxc_sliceimage and nodemanager
Thierry Parmentelat [Tue, 27 May 2014 09:23:01 +0000 (11:23 +0200)]
typo
Thierry Parmentelat [Tue, 27 May 2014 09:19:31 +0000 (11:19 +0200)]
remove band-aid code that was re-cleaning the slice rootfs before creating it
Thierry Parmentelat [Tue, 27 May 2014 09:16:17 +0000 (11:16 +0200)]
fixed nasty bug
Thierry Parmentelat [Fri, 23 May 2014 15:00:18 +0000 (17:00 +0200)]
fix typo in varname - umounts based on sliver name
Thierry Parmentelat [Fri, 23 May 2014 14:59:06 +0000 (16:59 +0200)]
do not call lsof, this is useless and produces huge logs
Thierry Parmentelat [Fri, 23 May 2014 14:36:29 +0000 (16:36 +0200)]
reviewd all timeouts for lxc slice creation/teradown
Thierry Parmentelat [Fri, 23 May 2014 14:34:44 +0000 (16:34 +0200)]
use logger.log_call to run mount/umount
Thierry Parmentelat [Fri, 23 May 2014 11:11:31 +0000 (13:11 +0200)]
patch for (hopefully) getting slice re-creation to work again
the sliver_lxc code for creating a sliver will check if /vservers/<slicename> exists, if so and it is empty nm will issue another call to
btrfs subvolume delete
in any case if the rootdir cannot be cleaned up sliver creation will be given up
Thierry Parmentelat [Wed, 21 May 2014 12:10:13 +0000 (14:10 +0200)]
giving up on packaging for ubuntu
Thierry Parmentelat [Wed, 21 May 2014 10:28:34 +0000 (12:28 +0200)]
deb pkg ctd
Thierry Parmentelat [Wed, 21 May 2014 10:08:35 +0000 (12:08 +0200)]
debian pkg
Thierry Parmentelat [Wed, 21 May 2014 09:31:31 +0000 (11:31 +0200)]
tentative fixes for debian
Thierry Parmentelat [Wed, 21 May 2014 08:21:38 +0000 (10:21 +0200)]
more aggressively try to cleanup sliver rootfs upon slice teardown
Thierry Parmentelat [Mon, 19 May 2014 14:07:28 +0000 (16:07 +0200)]
enable make sync to restart nm only conditionnally
Thierry Parmentelat [Fri, 16 May 2014 09:26:50 +0000 (11:26 +0200)]
typo
Thierry Parmentelat [Fri, 16 May 2014 08:38:59 +0000 (10:38 +0200)]
reviewed packaging contents as compared with the former scheme
in particular, was missing important initscript-related material
Thierry Parmentelat [Fri, 16 May 2014 06:42:48 +0000 (08:42 +0200)]
oops
Thierry Parmentelat [Thu, 15 May 2014 15:15:52 +0000 (17:15 +0200)]
remove unused Extension in setup
Thierry Parmentelat [Thu, 15 May 2014 11:39:29 +0000 (13:39 +0200)]
Merge branch 'packaging'
Thierry Parmentelat [Thu, 15 May 2014 10:36:00 +0000 (12:36 +0200)]
provide default to WITH_INIT and WITH_SYSTEMD right in the makefile
Thierry Parmentelat [Wed, 14 May 2014 14:29:48 +0000 (16:29 +0200)]
merged the 3 packaging-dependant tracks (-lib, -lxc and -vs) into a single specfile, single setup script and single Makefile
this is useful in its own, plus hopefully this will make ubuntu packaging much easier,
Thierry Parmentelat [Wed, 14 May 2014 12:29:14 +0000 (14:29 +0200)]
first and very rough attempt at building for debian
the way the 3 packages (lib, lxc, vs) are intermixed makes it hard under debian
I might want to rework the rpm packaging first
Thierry Parmentelat [Tue, 13 May 2014 16:41:49 +0000 (18:41 +0200)]
an attempt at robustifying slice teardown once again
not doing things in the exact right order causes /vservers/<slicename> to remain after slice deletion, thus preventing proper slice-re-creation
looks like finer grained test steps could help here some day
Thierry Parmentelat [Tue, 13 May 2014 09:34:39 +0000 (11:34 +0200)]
another bugfix
Thierry Parmentelat [Tue, 13 May 2014 09:22:08 +0000 (11:22 +0200)]
fix for ubuntus
Thierry Parmentelat [Tue, 13 May 2014 08:10:28 +0000 (10:10 +0200)]
add missing import subprocess
Thierry Parmentelat [Tue, 13 May 2014 08:10:10 +0000 (10:10 +0200)]
no change - reorder imports
Thierry Parmentelat [Tue, 13 May 2014 08:05:56 +0000 (10:05 +0200)]
added some comments about the fuse subsystem wrt systemd & vsys
Thierry Parmentelat [Mon, 28 Apr 2014 11:04:34 +0000 (13:04 +0200)]
Setting tag nodemanager-5.2-13
improve network management in slices
_has_systemctl was not declared as global - vsys was failing to restart
Scott Baker [Thu, 10 Apr 2014 08:07:13 +0000 (01:07 -0700)]
don't create network script if the slice hasn't been created yet
Scott Baker [Mon, 7 Apr 2014 22:33:37 +0000 (15:33 -0700)]
fix _has_systemctl referenced before assignment in python2.7
Scott Baker [Mon, 7 Apr 2014 21:13:48 +0000 (14:13 -0700)]
pass args to lxcsu separately when starting network