summary |
shortlog | log |
commit |
commitdiff |
tree
first ⋅ prev ⋅ next
Planet-Lab Support [Mon, 9 Jan 2006 21:41:04 +0000 (21:41 +0000)]
This commit was manufactured by cvs2svn to create tag
'planetlab-3_2_2'.
Mark Huang [Mon, 9 Jan 2006 21:41:02 +0000 (21:41 +0000)]
- merge revision 1.25
date: 2006/01/09 21:40:28; author: mlhuang; state: Exp; lines: +3 -0
- Some temporary cdroot files like /var/empty/sshd and
/usr/bin/sudo get created with non-readable permissions.
Mark Huang [Wed, 26 Oct 2005 17:54:59 +0000 (17:54 +0000)]
- merge revision 1.23
date: 2005/10/26 17:54:31; author: mlhuang; state: Exp; lines: +2 -2
- fix /var/run/vcached.pid generation
Planet-Lab Support [Sat, 1 Oct 2005 18:20:11 +0000 (18:20 +0000)]
This commit was manufactured by cvs2svn to create branch
'planetlab-3_2-branch'.
Mark Huang [Sat, 1 Oct 2005 18:20:08 +0000 (18:20 +0000)]
- use a valid node_id
Mark Huang [Wed, 14 Sep 2005 17:23:25 +0000 (17:23 +0000)]
- also copy /etc/planetlab/node_id
Mark Huang [Sat, 3 Sep 2005 21:49:03 +0000 (21:49 +0000)]
- execute new build script
- move most of %post script back to SysVInit script
Mark Huang [Sat, 3 Sep 2005 21:48:30 +0000 (21:48 +0000)]
- change back to SysVInit script that updates a few configuration files
in /vservers/vserver-reference from the host. Replicates a bit of the
functionality of Boot Manager.
Mark Huang [Sat, 3 Sep 2005 21:44:16 +0000 (21:44 +0000)]
Solve the bootstrap problem by including any just built packages in
the yum configuration. This cooperates with the PlanetLab build
system.
Mark Huang [Fri, 2 Sep 2005 19:46:37 +0000 (19:46 +0000)]
- get yum.conf from boot, not www
- use better check for whether RPMS exist to freshen
Mark Huang [Fri, 2 Sep 2005 15:21:00 +0000 (15:21 +0000)]
- be quiet in the curl request
Mark Huang [Fri, 2 Sep 2005 01:43:49 +0000 (01:43 +0000)]
- Get the URL for the production /etc/yum.conf file
Mark Huang [Thu, 1 Sep 2005 23:12:19 +0000 (23:12 +0000)]
- %post: update packages and populate header cache
Mark Huang [Thu, 1 Sep 2005 18:56:47 +0000 (18:56 +0000)]
- If run under sudo, allow user to delete the built RPM
- Stop vcached, delete the old reference image, and empty the cache
before installing a new version; restart vcached after installing
- make sure /vservers is mode 0000
- make sure the reference image has an /etc/planetlab directory before
trying to install in it
Mark Huang [Thu, 1 Sep 2005 18:55:11 +0000 (18:55 +0000)]
- add another comment about yum requirements
Mark Huang [Thu, 1 Sep 2005 17:51:26 +0000 (17:51 +0000)]
- chkconfig may not be in a regular user's PATH
Mark Huang [Thu, 1 Sep 2005 17:39:55 +0000 (17:39 +0000)]
- chkconfig may not be in a regular user's PATH
Mark Huang [Thu, 1 Sep 2005 17:36:54 +0000 (17:36 +0000)]
- If run under sudo, allow user to delete the build directory
Mark Huang [Thu, 1 Sep 2005 17:20:03 +0000 (17:20 +0000)]
- chroot may not be in a regular user's PATH
Mark Huang [Thu, 1 Sep 2005 17:17:14 +0000 (17:17 +0000)]
- exit with an error code if something fails
Mark Huang [Thu, 1 Sep 2005 17:08:50 +0000 (17:08 +0000)]
- Pre-package vserver-reference instead of building it on nodes
Mark Huang [Thu, 31 Mar 2005 22:34:42 +0000 (22:34 +0000)]
- /dev: make a symlink to /proc/self/fd so that bash process
substitution works
Mark Huang [Tue, 30 Nov 2004 18:24:02 +0000 (18:24 +0000)]
- bump release number
Mark Huang [Mon, 29 Nov 2004 20:19:13 +0000 (20:19 +0000)]
- PL3118 and PL3131 fix: set barrier bit on /vservers instead of old
immulink bit. Do not reset the immutable bit on the new
vserver-reference directory when deleting it after an error.
Mark Huang [Wed, 17 Nov 2004 22:54:19 +0000 (22:54 +0000)]
- back out previous checkin; getting weird errors if I do this on some
vuseradd invocations. Sigh.
Mark Huang [Wed, 17 Nov 2004 22:50:04 +0000 (22:50 +0000)]
- remove stale rpm locks from reference image after building
Mark Huang [Mon, 15 Nov 2004 19:27:49 +0000 (19:27 +0000)]
- bump release to install Fedora Core 2 updates as of Tue Nov 9 2004
- PL3017 fix: rebuild vserver-reference image in case it was built
with i386 glibc
Mark Huang [Thu, 4 Nov 2004 18:11:59 +0000 (18:11 +0000)]
bump release number
Mark Huang [Fri, 29 Oct 2004 20:29:13 +0000 (20:29 +0000)]
- write a fake mtab so that df works
Mark Huang [Wed, 27 Oct 2004 15:11:41 +0000 (15:11 +0000)]
change URL to repository location
Mark Huang [Wed, 20 Oct 2004 19:55:41 +0000 (19:55 +0000)]
- dump stdout/stderr to a log file /var/log/vserver-reference.log so we
can debug when vserver-reference init fails
Mark Huang [Thu, 14 Oct 2004 18:37:40 +0000 (18:37 +0000)]
- start earlier (after ssh but before pl_conf, pl_nm, vcached, and
vservers)
Mark Huang [Thu, 14 Oct 2004 18:25:05 +0000 (18:25 +0000)]
- move functionality into /etc/init.d/vserver-reference, which runs once
after installation (and maybe again if the RPM is ever updated, but the
script is mostly idempotent).
Mark Huang [Wed, 13 Oct 2004 04:53:07 +0000 (04:53 +0000)]
- PL2437 fix: build new vserver-reference image in /vservers/.vtmp, then
swap with /vservers/vserver-reference only when done. Bail and clean
up on any error.
Mark Huang [Mon, 11 Oct 2004 04:25:17 +0000 (04:25 +0000)]
reference image now dynamically built
Mark Huang [Mon, 11 Oct 2004 04:24:42 +0000 (04:24 +0000)]
- dynamically construct reference image on the target machine in a %post
scriplet, instead of building it on the build machine. This solves a
couple of problems:
1. RPM databases constructed on a RH9 machine could be corrupt on a
FC2 machine because of the LD_ASSUME_KERNEL rpm/glibc NPTL bug.
2. vserver-reference is no longer a 50 MB download. The reference
image can be installed from the localhost yum cache most of the
time.
But introduces a couple of problems:
1. If the %post scriplet fails, the whole installation fails
with a warning, but without an error. Not having a complete
vserver-reference build is a bad thing.
2. Must zap RPM and YUM locks while installing.
Mark Huang [Sun, 10 Oct 2004 23:17:51 +0000 (23:17 +0000)]
- escape % correctly
Mark Huang [Thu, 7 Oct 2004 17:10:06 +0000 (17:10 +0000)]
- RPM is too $#%@! stupid to understand bash here documents and tries
too hard to expand macros
Mark Huang [Thu, 7 Oct 2004 16:59:24 +0000 (16:59 +0000)]
- set __file_context_path to /dev/null avoid stupid SELinux warning
Mark Huang [Tue, 5 Oct 2004 14:58:15 +0000 (14:58 +0000)]
- bootstrap: freshen any RPMs within the VServer reference image, with
RPMs that just got built at the same time as the vserver-reference RPM
itself
Mark Huang [Thu, 23 Sep 2004 15:10:54 +0000 (15:10 +0000)]
- exclude docs installation by default
Mark Huang [Mon, 20 Sep 2004 05:12:27 +0000 (05:12 +0000)]
- sudo: make sure the original user can remove the generated files
Mark Huang [Sat, 18 Sep 2004 21:09:40 +0000 (21:09 +0000)]
- don't install alpha RPMs in the vserver-reference image
Mark Huang [Sat, 18 Sep 2004 20:46:10 +0000 (20:46 +0000)]
- don't build debuginfo package
Mark Huang [Sat, 18 Sep 2004 20:22:35 +0000 (20:22 +0000)]
- add /sbin and /usr/sbin to our PATH in case we are running in an sudo
environment
Mark Huang [Sat, 18 Sep 2004 20:22:13 +0000 (20:22 +0000)]
- add a dummy planetlab-v3/ source for now to get yumgroups.xml for v3
Mark Huang [Sat, 18 Sep 2004 19:44:37 +0000 (19:44 +0000)]
- use our own yum.conf
- XXX should get it from PlanetLabConf instead
Mark Huang [Sat, 18 Sep 2004 19:36:36 +0000 (19:36 +0000)]
This commit was generated by cvs2svn to compensate for changes in r210,
which included commits to RCS files with non-trunk default branches.
Planet-Lab Support [Wed, 4 Jun 2003 20:07:28 +0000 (20:07 +0000)]
New repository initialized by cvs2svn.