Faiyaz Ahmed [Thu, 6 Nov 2008 22:26:03 +0000 (22:26 +0000)]
Rough draft to support admin tickets from "blessed" interfaces.
Faiyaz Ahmed [Wed, 8 Oct 2008 14:55:42 +0000 (14:55 +0000)]
Syntax error
Faiyaz Ahmed [Mon, 6 Oct 2008 16:29:14 +0000 (16:29 +0000)]
Untested
Faiyaz Ahmed [Mon, 22 Sep 2008 17:41:50 +0000 (17:41 +0000)]
Another optimization. Bringing us closer to generic support for cliques
Faiyaz Ahmed [Fri, 19 Sep 2008 19:06:20 +0000 (19:06 +0000)]
Check for the existence of the Internet2 clique iptables rule and add if absent.
Faiyaz Ahmed [Wed, 17 Sep 2008 18:24:34 +0000 (18:24 +0000)]
Clique network shares don't exist yet.
Faiyaz Ahmed [Wed, 10 Sep 2008 20:34:33 +0000 (20:34 +0000)]
Move checks from bwmon for running limits before calling bwlimit.
Faiyaz Ahmed [Tue, 9 Sep 2008 16:47:50 +0000 (16:47 +0000)]
* Explicitly run configure, which sets capabilities and resources, before starting vserver.
* Print username when NM api returns permission denied.
Faiyaz Ahmed [Wed, 3 Sep 2008 20:57:20 +0000 (20:57 +0000)]
enable disable flag.
Thierry Parmentelat [Mon, 1 Sep 2008 15:04:27 +0000 (15:04 +0000)]
from branch
Faiyaz Ahmed [Mon, 25 Aug 2008 19:03:35 +0000 (19:03 +0000)]
Fix toggling of email flag.
Faiyaz Ahmed [Thu, 21 Aug 2008 16:43:10 +0000 (16:43 +0000)]
bwmon: Fixed extra notification when 1 of 2 limits weren't exceeded.
Faiyaz Ahmed [Thu, 21 Aug 2008 02:44:20 +0000 (02:44 +0000)]
Fixed key error in logging.
Daniel Hokka Zakrisson [Thu, 21 Aug 2008 01:23:47 +0000 (01:23 +0000)]
If the kernel doesn't have support for fuse, exit nicely.
Faiyaz Ahmed [Wed, 20 Aug 2008 23:33:32 +0000 (23:33 +0000)]
clean up message template and logging. syntax errors.
Faiyaz Ahmed [Wed, 20 Aug 2008 23:20:30 +0000 (23:20 +0000)]
syntax error. fixed.
Faiyaz Ahmed [Wed, 20 Aug 2008 23:14:56 +0000 (23:14 +0000)]
syntax error. fixed.
Faiyaz Ahmed [Mon, 18 Aug 2008 19:27:14 +0000 (19:27 +0000)]
text.
Daniel Hokka Zakrisson [Sun, 17 Aug 2008 22:41:37 +0000 (22:41 +0000)]
Expand the glob.
Thierry Parmentelat [Thu, 14 Aug 2008 06:00:30 +0000 (06:00 +0000)]
fix build
Daniel Hokka Zakrisson [Wed, 13 Aug 2008 15:45:55 +0000 (15:45 +0000)]
FUSE patch from Jeremy Stribling.
Faiyaz Ahmed [Tue, 12 Aug 2008 19:16:58 +0000 (19:16 +0000)]
Round to ints
Faiyaz Ahmed [Tue, 12 Aug 2008 19:02:37 +0000 (19:02 +0000)]
Threads with queues using global timers for throttling execution rates are a BAD IDEAsvn diff
Faiyaz Ahmed [Mon, 11 Aug 2008 18:34:45 +0000 (18:34 +0000)]
NM has a worker thread for every sliver ever to exist on the node. Each thread works off of a queue where the database adds a state from the statemachine to execute. To keep from spawning upwards of 300 worker processes during start(), a cummulative global delay is incremented (by 2) for every sliver known to the db. After the delay, start() is run. But, it's possible for the delay to grow without bound when NM is resynced to PLC and the db adds previously added states to the queue and causing the cumm start delay to be incremented once again. This patch keeps the queue from growing beyond the 4 known states. Any other additions are likely repeates and don't need to be added again. This is a bandaid; really we should keep a table of what exists in the queue and keep from readding while also bounding the delay to some maximum over the number of slivers present on the node.
Faiyaz Ahmed [Fri, 1 Aug 2008 16:36:00 +0000 (16:36 +0000)]
2 way merge from branch. Includes support for local_ scripts.
Faiyaz Ahmed [Fri, 1 Aug 2008 16:10:49 +0000 (16:10 +0000)]
Get defaults from bwlimit. Also remove checks before issuing bwlimit commands. Bwlimit now checks running values before committing new ones.
Faiyaz Ahmed [Mon, 28 Jul 2008 19:45:40 +0000 (19:45 +0000)]
Chown .ssh to user:slices. SSH will bail otherwise.
Faiyaz Ahmed [Sat, 26 Jul 2008 02:47:44 +0000 (02:47 +0000)]
change ownership of authorized_keys.
Faiyaz Ahmed [Tue, 22 Jul 2008 22:35:37 +0000 (22:35 +0000)]
* Slices w/ initscripts were starting repeatedly because the flag that signals the initscript has not changed was being toggled after the fork(). Fixed.
Faiyaz Ahmed [Mon, 21 Jul 2008 23:55:39 +0000 (23:55 +0000)]
There's no need to fork to write to a file.
Faiyaz Ahmed [Tue, 15 Jul 2008 21:10:56 +0000 (21:10 +0000)]
Added logging. Forgot to return anything; threads would blow up since test of is_runing would fail.
Faiyaz Ahmed [Mon, 14 Jul 2008 22:00:52 +0000 (22:00 +0000)]
Add check to see if sliver is_running() during ensure_created. Previously, if the object existed in the database, it was assumed the sliver was running and the start() operation would ignore the record. However, if a sliver is damaged, its best to query util-vserver to actually make sure the slice is indeed running.
Faiyaz Ahmed [Tue, 8 Jul 2008 17:46:08 +0000 (17:46 +0000)]
Disregard scripts that start with "local_".
Daniel Hokka Zakrisson [Mon, 7 Jul 2008 16:13:22 +0000 (16:13 +0000)]
Use the right variable.
Daniel Hokka Zakrisson [Mon, 7 Jul 2008 16:02:42 +0000 (16:02 +0000)]
chrooting is not thread-safe. Install the initscript before calling VServer.start instead, where we know we're the only thread running.
Faiyaz Ahmed [Fri, 27 Jun 2008 16:46:12 +0000 (16:46 +0000)]
Print log only after global vdu semaphore is acquired. Misleading otherwise.
Thierry Parmentelat [Tue, 24 Jun 2008 07:46:18 +0000 (07:46 +0000)]
applied changeset 9676 on trunk
Faiyaz Ahmed [Fri, 20 Jun 2008 01:17:40 +0000 (01:17 +0000)]
this is what i really wanted to do. temporary fix. this needs to be cleaned up
Faiyaz Ahmed [Fri, 20 Jun 2008 00:58:03 +0000 (00:58 +0000)]
Should be able to explicitly specify as well.
Faiyaz Ahmed [Tue, 17 Jun 2008 18:47:22 +0000 (18:47 +0000)]
Timeout curl in the event Apache (PLC side) opens the XMLRPC session, but returns no data for specified timeout.
Faiyaz Ahmed [Tue, 17 Jun 2008 18:43:49 +0000 (18:43 +0000)]
5 minutes is too long of a timeout to keep an apache proc open on PLC. 90 seconds is probably more than enough.
Faiyaz Ahmed [Tue, 17 Jun 2008 18:40:45 +0000 (18:40 +0000)]
for exception handling.
Thierry Parmentelat [Mon, 9 Jun 2008 15:24:29 +0000 (15:24 +0000)]
taglevel on the 1.7-11 tag had remained 10 - also moving the trunk to 1.8-0 as 1.7 is confusing
Faiyaz Ahmed [Wed, 4 Jun 2008 20:42:38 +0000 (20:42 +0000)]
Log when talking to PLC
Faiyaz Ahmed [Wed, 4 Jun 2008 19:58:38 +0000 (19:58 +0000)]
RIP Proper.
Thierry Parmentelat [Tue, 3 Jun 2008 08:12:35 +0000 (08:12 +0000)]
5.0
Faiyaz Ahmed [Fri, 30 May 2008 19:50:17 +0000 (19:50 +0000)]
Removed unnecessary comment.
Thierry Parmentelat [Fri, 30 May 2008 14:03:07 +0000 (14:03 +0000)]
cosmetic
Faiyaz Ahmed [Thu, 29 May 2008 17:26:36 +0000 (17:26 +0000)]
Do not wait for slices to exist before adding to vsys config and acls.
Thierry Parmentelat [Wed, 28 May 2008 15:49:13 +0000 (15:49 +0000)]
fixes
Thierry Parmentelat [Wed, 28 May 2008 10:22:40 +0000 (10:22 +0000)]
get rid of curlwrapper.CurlException and raise xmlrpclib.ProtocolError instead
temporarily prints as many details as possible when curl fails
Thierry Parmentelat [Wed, 28 May 2008 07:15:56 +0000 (07:15 +0000)]
expose more info when certificate check fails
Thierry Parmentelat [Mon, 26 May 2008 14:13:19 +0000 (14:13 +0000)]
moving towards 5.0
Thierry Parmentelat [Wed, 14 May 2008 01:44:39 +0000 (01:44 +0000)]
myplc-docs build from local dtds - more robust, and much faster too
Faiyaz Ahmed [Fri, 9 May 2008 21:18:31 +0000 (21:18 +0000)]
Removed comments.
Thierry Parmentelat [Fri, 9 May 2008 06:38:36 +0000 (06:38 +0000)]
fix3
Faiyaz Ahmed [Thu, 8 May 2008 21:11:32 +0000 (21:11 +0000)]
Revert to forking before calling VServer.start() to avoid creating zombies.
Thierry Parmentelat [Thu, 8 May 2008 09:11:02 +0000 (09:11 +0000)]
first draft for the myplc-docs package (cross-module change)
Thierry Parmentelat [Thu, 24 Apr 2008 11:38:06 +0000 (11:38 +0000)]
add f9 to the list of known pldistros
Faiyaz Ahmed [Wed, 16 Apr 2008 21:26:25 +0000 (21:26 +0000)]
setname as part of set_resources. Better than stupid sleep while part of start(). Called from VServer.__do_chcontext().
Faiyaz Ahmed [Wed, 16 Apr 2008 18:51:00 +0000 (18:51 +0000)]
Setname failed because context wasn't setup. Added busy wait in VServer.is_running() for 5 minutes before setting name. Also removed first fork before calling VServer.start(). Moved fd closing code to util-vserver to avoid zombies.
Faiyaz Ahmed [Mon, 14 Apr 2008 21:58:13 +0000 (21:58 +0000)]
Set slice_id as vcVHI_CONTEXT.
Faiyaz Ahmed [Wed, 9 Apr 2008 18:38:53 +0000 (18:38 +0000)]
BW limits and resources are handled by bwmon.py
Faiyaz Ahmed [Wed, 9 Apr 2008 18:08:45 +0000 (18:08 +0000)]
Use API host when Planetflow host isn't defined.
Faiyaz Ahmed [Fri, 4 Apr 2008 19:46:46 +0000 (19:46 +0000)]
Debugging output.
Faiyaz Ahmed [Fri, 4 Apr 2008 19:44:34 +0000 (19:44 +0000)]
File locking isn't exclusive in the same process across threads. Switched to regular semaphores. The assumption is sliver_vs.configure is called before Vserver.start(), which sets disk_usage_initialized. The disk check should be skipped during start, so the lock won't carry over. ...I hope.
Faiyaz Ahmed [Thu, 3 Apr 2008 22:15:19 +0000 (22:15 +0000)]
Fix config file parsing. Blank lines were messing thigns up.
Faiyaz Ahmed [Wed, 2 Apr 2008 21:12:51 +0000 (21:12 +0000)]
Updated changelog
Faiyaz Ahmed [Wed, 2 Apr 2008 21:02:18 +0000 (21:02 +0000)]
Python2.5 changed the max levels of inheritence.
Faiyaz Ahmed [Wed, 2 Apr 2008 19:07:45 +0000 (19:07 +0000)]
Added support for multiple host -> same slice mappings
Faiyaz Ahmed [Tue, 1 Apr 2008 21:47:59 +0000 (21:47 +0000)]
Bump revision.
Faiyaz Ahmed [Tue, 1 Apr 2008 21:39:17 +0000 (21:39 +0000)]
Syntax error when writing config file
Faiyaz Ahmed [Fri, 28 Mar 2008 17:50:54 +0000 (17:50 +0000)]
Bump versions
Faiyaz Ahmed [Thu, 27 Mar 2008 20:34:26 +0000 (20:34 +0000)]
get config from file.
Faiyaz Ahmed [Thu, 27 Mar 2008 19:17:51 +0000 (19:17 +0000)]
sort domain name. lower order domains should be first.
Faiyaz Ahmed [Tue, 25 Mar 2008 19:21:16 +0000 (19:21 +0000)]
root slice needs to come first. Also, emtpy conf file now handled correctly.
Faiyaz Ahmed [Tue, 25 Mar 2008 17:35:01 +0000 (17:35 +0000)]
* Checks for zombied pids during restart
* Fixed toggling of instantiated slices.
Faiyaz Ahmed [Mon, 24 Mar 2008 21:41:20 +0000 (21:41 +0000)]
add carriage return to lines
Faiyaz Ahmed [Mon, 24 Mar 2008 21:30:00 +0000 (21:30 +0000)]
Added codemux
Faiyaz Ahmed [Mon, 24 Mar 2008 21:29:32 +0000 (21:29 +0000)]
* Now callback in NM by default
* Clean up after slice is removed
Faiyaz Ahmed [Mon, 24 Mar 2008 21:00:32 +0000 (21:00 +0000)]
path fix.
Faiyaz Ahmed [Mon, 24 Mar 2008 20:00:59 +0000 (20:00 +0000)]
Initial Commit.
Faiyaz Ahmed [Mon, 24 Mar 2008 19:50:38 +0000 (19:50 +0000)]
Initial commit.
Thierry Parmentelat [Sat, 22 Mar 2008 11:03:45 +0000 (11:03 +0000)]
multiple slice falvours - second iteration
* review naming scheme for system slices; now planetflow vref image is named like
planetflow-f8-i386 for consistency
* use for sensible name for the stamp where default slicefamily is stored
Thierry Parmentelat [Fri, 21 Mar 2008 21:09:39 +0000 (21:09 +0000)]
fixed a couple remaining issues, seems to work
Thierry Parmentelat [Fri, 21 Mar 2008 20:55:48 +0000 (20:55 +0000)]
fixed syntax error
Faiyaz Ahmed [Fri, 21 Mar 2008 18:27:10 +0000 (18:27 +0000)]
Check to see if slice is running first before creating /vsys directory.
Thierry Parmentelat [Fri, 21 Mar 2008 16:35:32 +0000 (16:35 +0000)]
support for multiple slice reference images
* vserver-reference rpm names have changed and become
vserver-@pldistro@-@fcdistro@-@arch@
vserver-systemslices-@pldistro@-@fcdistro@-@arch@
* this contains the images located as
/vservers/.vref/@pldistro@-@fcdistro@-@arch@
and
/vservers/.vstub/@pldistro@-@fcdistro@-@arch@/planetflow
respectively
* a stamp gets created on the nodes as /etc/planetlab/defaultvref
that contains the name for the default slicefamily to use
* nodemanager tries to be smart when chosing the right argument to pass to vuseradd
for this purpose, the argument in 'vref' is taken as a dash-separated list of 'wishes'
so e.g. is defaultvref is planetlab-f8-x86_64, then we should find for the image name:
vref=i386 -> vuseradd -t planetlab-f8-i386
vref=centos5 -> vuseradd -t planetlab-centos5-x86_64
vref=planetflow -> vuseradd -t planetflow
vref=centos5-onelab -> vuseradd -t onelab-centos5-x86_64
for legacy reasons, if the stamp is not found, 'default' is taken instead
=== status
* this is only partly tested
* a given node can install several variants of vserver-*
this would require a manual upload of the rpms into the right install-rpms
and manual changes in yumgroups.
* next improvement would be to have the system (planetflow) slices have a smarter naming scheme as well
this way a 'planetflow' vref would rather be rewritten as planetflow-f8-x86_64
for the time being it is not supported to install several variants of vserver-systemslices
on the node - although it should be harmless
Faiyaz Ahmed [Fri, 7 Mar 2008 22:24:31 +0000 (22:24 +0000)]
Bug fix for caching dead slices.
Faiyaz Ahmed [Fri, 7 Mar 2008 20:34:59 +0000 (20:34 +0000)]
Less noisey logging. Also uses LOG_LEVEL now.
Faiyaz Ahmed [Thu, 6 Mar 2008 22:06:13 +0000 (22:06 +0000)]
fixed deletion logic. ..this is sloppy but it works. Also fixed network shares.
Faiyaz Ahmed [Thu, 6 Mar 2008 16:01:56 +0000 (16:01 +0000)]
to support pl_mom list emails instead of mails to support
John Doe [Wed, 5 Mar 2008 09:43:05 +0000 (09:43 +0000)]
simple patch for redirecting pl-mom notification to the right mailing list
Stephen Soltesz [Tue, 4 Mar 2008 19:26:17 +0000 (19:26 +0000)]
Added code to reduce duplicate entries in /etc/vsys.conf; this will help play
more nicely with vsys.
Stephen Soltesz [Fri, 29 Feb 2008 19:27:55 +0000 (19:27 +0000)]
Added stuff for the NM API.
Stephen Soltesz [Fri, 29 Feb 2008 19:27:12 +0000 (19:27 +0000)]
Script to automatically upload changes in a drupal page to drupal. Handy for
the api. This almost works, but not quite. It worked on a recent myplc, but doesn't
work on the production box. I don't know why.
Faiyaz Ahmed [Wed, 27 Feb 2008 14:28:42 +0000 (14:28 +0000)]
Added sharing for loans.
Thierry Parmentelat [Thu, 14 Feb 2008 18:33:10 +0000 (18:33 +0000)]
Configures vsys via vsys slice attribute {name: vsys, value: script}
CPU reservations are now calculated via percentages instead of shares
BW totals preserved for dynamic slices
Closes bug where node cap sets off bw slice alarms for all slices.
Manual fix after a broken module-tag.py attempt
Faiyaz Ahmed [Wed, 13 Feb 2008 17:25:25 +0000 (17:25 +0000)]
Include vsys.
Faiyaz Ahmed [Wed, 13 Feb 2008 17:24:46 +0000 (17:24 +0000)]
include vsys call back.