Ryota Ozaki wrote: [Tue Oct 13 2009, 05:05:15PM EDT]
Hi,
lxc controller usually executes 'ip' command to control veths, however
now PATH environment variable of the controller is not set (logging
code resets environment variables all) and the execution will definitely
fail because the controller
Daniel P. Berrange wrote: [Tue Oct 06 2009, 05:00:55AM EDT]
Having looked at the wway the next patch uses these, I think it'd be
nicer to change the contract to just be
extern char *virLogGetFilters(void);
extern char *virLogGetOutputs(void);
Heh, that's how I wrote it
Daniel P. Berrange wrote: [Mon Oct 05 2009, 08:52:20AM EDT]
ACK, although it'd need a small change for my suggested API contract
change in your previous patch in the series
Updated for API change...
Before launching the lxc controller, have the lxc driver query the log
settings and setup
Updated for API change.
Add a config file for the lxc driver, based on existing qemu.conf.
There is currently one tunable log_with_libvirtd that controls
whether an lxc controller will log only to the container log file, or
whether it will honor libvirtd's log output configuration. This
provides
Paolo Bonzini wrote: [Mon Oct 05 2009, 04:54:27AM EDT]
On 10/04/2009 09:28 PM, Amy Griffis wrote:
Add a utility to ensure an absolute path for a potentially realtive path.
Would it make sense for your usage to resolve symbolic links at the
same time? If so, you can use the canonicalize
Daniel P. Berrange wrote: [Mon Oct 05 2009, 08:51:00AM EDT]
On Sun, Oct 04, 2009 at 03:28:46PM -0400, Amy Griffis wrote:
When configuring logging settings, keep more information about the
output destination. Add accessors to retrieve the filter and output
settings in the original string
The following patches fix logging for the lxc controller. The lxc
controller makes use of the libvirt logging module, but because it
doesn't re-initialize the logging configuration the messages are
currently going nowhere.
There are a couple of options in fixing this. The lxc controller could
use
Add a utility to ensure an absolute path for a potentially realtive path.
---
src/libvirt_private.syms |1 +
src/util/util.c | 36
src/util/util.h |3 +++
3 files changed, 40 insertions(+), 0 deletions(-)
diff --git
When configuring logging settings, keep more information about the
output destination. Add accessors to retrieve the filter and output
settings in the original string form; this to be used to set up
environment for a child process that also logs. Open output files
O_APPEND so child can also write
Some callers may have set envp[].
---
src/util/util.c | 13 -
1 files changed, 12 insertions(+), 1 deletions(-)
diff --git a/src/util/util.c b/src/util/util.c
index 81b743c..e5135fc 100644
--- a/src/util/util.c
+++ b/src/util/util.c
@@ -601,12 +601,23 @@
Do we really want to overwrite the container log file every time we
restart?
---
src/lxc/lxc_driver.c |2 +-
1 files changed, 1 insertions(+), 1 deletions(-)
diff --git a/src/lxc/lxc_driver.c b/src/lxc/lxc_driver.c
index c5a49f2..73cfecd 100644
--- a/src/lxc/lxc_driver.c
+++
Add a config file for the lxc driver, based on existing qemu.conf.
There is currently one tunable log_with_libvirtd that controls
whether an lxc controller will log only to the container log file, or
whether it will honor libvirtd's log output configuration. This
provides a way to have libvirtd
I've attempted to add augeas support for lxc.conf based on qemu.conf.
---
libvirt.spec.in | 10 ++
src/Makefile.am | 18 --
src/lxc/libvirtd_lxc.aug | 30 ++
src/lxc/test_libvirtd_lxc.aug | 31
Mark McLoughlin wrote: [Thu Aug 06 2009, 08:25:06AM EDT]
On Fri, 2009-07-31 at 17:57 -0400, Amy Griffis wrote:
+pLibvirtd does not reload its logging configuration when issued a
SIGHUP.
+ If you want to reload the configuration, you must do a codeservice
+ libvirtd
Daniel P. Berrange wrote: [Thu Aug 06 2009, 10:07:26AM EDT]
I guess that should be service libvirtd restart ?
(i.e. all reload does is issue a SIGHUP)
Oh, I forgot I was looking at debian. :-) Their start script for
libvirtd actually does a restart under the reload. Generalizing
The following patches fixup libvirt's logging code, to hopefully make it
a little more consistent and predictable.
---
Amy Griffis (5):
Tighten libvirt's parsing of logging environment variables
Several fixes to libvirtd's log setup
Cleanup VIR_LOG_DEBUG parsing in eventtest
Don't convert high priority levels to the debug level. Don't parse
LIBVIRT_LOG_FILTERS and LIBVIRT_LOG_OUTPUTS when they're set to the
empty string. Warn when the user specifies an invalid value (empty
string remains a noop).
---
src/libvirt.c | 13 -
src/logging.c | 52
Similar as for general libvirt, don't convert high priority levels to debug
level. Ignore LIBVIRT_LOG_FILTERS and LIBVIRT_LOG_OUTPUTS when they're set to
the empty string, otherwise they can override a valid setting from the config
file. Send all settings through the parser functions for
Don't covert high priority levels to debug level. Consider an invalid priority
level setting a setup failure.
---
tests/eventtest.c | 12
1 files changed, 8 insertions(+), 4 deletions(-)
diff --git a/tests/eventtest.c b/tests/eventtest.c
index ff74b04..05fe3f3 100644
---
Add two new functions to the internal API,
virLogParseDefaultPriority() and virLogSetFromEnv(),
as was suggested earlier by Cole Robinson.
---
qemud/qemud.c| 50 ++
src/libvirt.c| 22 +---
Try to include a little more description about the corner cases,
things someone might get hung up on on.
---
docs/logging.html.in | 22 --
1 files changed, 20 insertions(+), 2 deletions(-)
diff --git a/docs/logging.html.in b/docs/logging.html.in
index fcd100f..1f3085d
Some more logging cleanup. Remove all references to log level 0,
and replace with VIR_LOG_DEBUG where appropriate. Also change any
hard numbers to the corresponding VirLogPriority and add
VIR_LOG_DEFAULT for the starting log level.
Signed-off-by: Amy Griffis amy.grif...@hp.com
---
docs
section since either setting is effectively treated
the same way in the following code.
Signed-off-by: Amy Griffis amy.grif...@hp.com
diff --git a/src/Makefile.am b/src/Makefile.am
index f65e7ad..23a1843 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -498,10 +498,6 @@ EXTRA_DIST
actually turns off
the entire logging module instead of just the debug output. I can
work on fixing that part in another patch.
Signed-off-by: Amy Griffis amy.grif...@hp.com
diff --git a/src/uml_driver.c b/src/uml_driver.c
index 7949d4f..7aeb7bf 100644
--- a/src/uml_driver.c
+++ b/src/uml_driver.c
Taking a second look at the log level code and documentation,
there seems to be some confusion around the value of '0'. It's
not one of the defined log priorities, but it is mentioned in the
documentation on the website both as meaning log everything and
no logging at all.
In the code, there are
Daniel P. Berrange wrote: [Wed Jun 17 2009, 06:44:31AM EDT]
On Tue, Jun 16, 2009 at 06:12:04PM -0400, Amy Griffis wrote:
Cole Robinson wrote: [Tue Jun 16 2009, 02:44:28PM EDT]
On 06/16/2009 01:35 PM, Amy Griffis wrote:
The lxc controller can't see libvirtd's log level setting so
The lxc controller can't see libvirtd's log level setting so it
needs to re-query it from the environment. The parsing code has
a few users now, so I added a new function to the internal API,
virLogParseDefaultPriority() along the lines of the other parse
functions.
Signed-off-by: Amy Griffis
Cole Robinson wrote: [Tue Jun 16 2009, 02:44:28PM EDT]
On 06/16/2009 01:35 PM, Amy Griffis wrote:
The lxc controller can't see libvirtd's log level setting so it
needs to re-query it from the environment. The parsing code has
a few users now, so I added a new function to the internal API
Update lxcDomainUndefine() to use VIR_ERR_INVALID_OPERATION
instead of VIR_ERR_INTERNAL_ERROR.
Signed-off-by: Amy Griffis amy.grif...@hp.com
diff --git a/src/lxc_driver.c b/src/lxc_driver.c
index 5392af1..64c735d 100644
--- a/src/lxc_driver.c
+++ b/src/lxc_driver.c
@@ -329,13 +329,13 @@ static
29 matches
Mail list logo