summary |
shortlog | log |
commit |
commitdiff |
tree
first ⋅ prev ⋅ next
root [Fri, 27 May 2005 06:41:43 +0000 (08:41 +0200)]
[multipathd] don't switch to already active group
This simple protection saves us a DM event per switch_group message.
root [Thu, 26 May 2005 21:35:45 +0000 (23:35 +0200)]
[multipathd] failback property selector fixes
Cut'n paste haste ... corrected
Should fix Tran Lan report added to bug [156280] today.
root [Wed, 25 May 2005 09:53:18 +0000 (11:53 +0200)]
[multipathd] add failback keyword handling
The failback keyword can be set in the default section, or in a hardware
entry, or in a multipath entry. Its value can be "manual", "immediate" or
an integer delay expressed in seconds.
root [Tue, 24 May 2005 09:30:25 +0000 (11:30 +0200)]
[PATCH][Bug 158603] local variable used before set
.
root [Tue, 17 May 2005 14:52:17 +0000 (16:52 +0200)]
[multipathd]
Complete the uevent framework.
The missing bit was the devmap add/remove events handling. Now that it
is in place, get rid of wait_thr altogether : kill the waiters vector,
the "lease" thing.
Helpers added in the process :
o libmultipath/devmapper.c:dm_mapname()
o libmultipath/structs.c:find_mp_by_minor()
Worth noting, devmap remove uevents don't get reliably delivered.
If someone have an explanation I'm interested.
root [Mon, 16 May 2005 21:45:49 +0000 (23:45 +0200)]
[libmultipath]
uevents delivery is too fast for sysfs files creation to complete
before we try to use them in discovery.c
The problem is well known to udev developpers, so merge their
wait loop in discovery.c
root [Mon, 16 May 2005 21:08:31 +0000 (23:08 +0200)]
[PATCH] call dm_lib_release/exit explicitly in multipath-tools
multipath-tools should call dm_lib_release() and dm_lib_exit()
before it returns. Then, libdevmapper will call update_devs()
and finalize the device node operation which creates
/dev/mapper/<map_name>.
This patch may not be needed if the latest libdevmapper of the
device-mapper CVS is used, since in the latest version,
dm_lib_exit() calls dm_lib_release() and dm_lib_exit() is called
by default when the program returns.
Regards,
Kiyoshi Ueda
root [Mon, 16 May 2005 21:04:53 +0000 (23:04 +0200)]
[multipathd]
When a path is added through a uevent, set the pp->mpp ownership.
A little infrastructure added to help :
o store_pathinfo() now returns the path pointer
o add find_mp_by_wwid()
root [Mon, 16 May 2005 13:08:19 +0000 (15:08 +0200)]
[multipathd]
Jumbo patch :
o get rid of the last multipath configurator call
o introduce the native path group switching logic.
For now only two behaviours :
o FAILBACK_MANUAL
o FAILBACK_IMMEDIATE
o get rid of path group ids unused bits
o paths and multipaths full scan is only done once now.
Moved one step earlier, in child(). Lists housekeeping done
through uevents (well, the mpvec part is yet to come)
root [Fri, 13 May 2005 11:56:18 +0000 (13:56 +0200)]
[multipathd]
Reinstate paths from the daemon. Don't rely on successful multipath exec
for that anymore.
Multipath is now only used to do the path group switches, which will
eventualy be merge in the daemon too.
Plug a locking weakness introduced with pp->mpp usage.
root [Fri, 13 May 2005 09:06:32 +0000 (11:06 +0200)]
[config]
Comment all lines in the example configuration file to force users to
uncomment only what they need.
This best practice was suggested by Kiyoshi Ueda, NEC.
root [Thu, 12 May 2005 18:15:53 +0000 (20:15 +0200)]
[multipathd]
The "checking paths" debug message is not correct anymore. Change that to
a nice "tick" :/
root [Thu, 12 May 2005 18:13:17 +0000 (20:13 +0200)]
[multipathd]
The load_config() introduction in the daemon broke the polling_interval
keyword parsing. Fix That.
root [Thu, 12 May 2005 18:05:26 +0000 (20:05 +0200)]
[multipathd]
Checker delay throttle, the easy way :
o max_checkint is set to (checkint * 4)
o after one successful checks, each consecutive successful check bump
up the checkint for the path by a factor of 2, toping at max_checkint
o path state change and mark_failed_paths() reset the checkint to its
minimal value
root [Thu, 12 May 2005 16:49:36 +0000 (18:49 +0200)]
[kpartx]
Try another approach at fixing the dev_t definition glitch with glibc headers
root [Thu, 12 May 2005 09:23:36 +0000 (11:23 +0200)]
[multipathd]
Refresh devmap list and initiate DM event waiter threads housekeeping upon device map add/remove uevent.
root [Mon, 9 May 2005 14:11:40 +0000 (16:11 +0200)]
[multipathd]
Remove unsupported getopt() flags, mistakenly inherited from multipath
root [Mon, 9 May 2005 14:10:05 +0000 (16:10 +0200)]
[multipathd]
Use load_config() in the daemon too.
root [Mon, 9 May 2005 14:01:38 +0000 (16:01 +0200)]
[libmultipath]
In load_config(), don't bother allocating a new struct config if already allocated. Don't override a previously set conf->verbosity (by getopt for example)
root [Mon, 9 May 2005 13:57:18 +0000 (15:57 +0200)]
[libmultipath]
Remove a signal remnent in struct config and load_config()
root [Thu, 5 May 2005 22:07:16 +0000 (00:07 +0200)]
[multipathd]
Logging code in multipathd cause unaligned access
which, on ia64, yields logs on both console and syslog like:
unaligned access to 0x60000000000141d1, ip=0x4000000000009d81
It's harmless except for the slight performance effect.
However, it's annoying for users and better to fix.
Attached patch aligns the la->tail to pointer size.
Jun'ichi Nomura, NEC
root [Wed, 4 May 2005 15:46:28 +0000 (17:46 +0200)]
[multipath]
0:0:0:0 is a valid host/bus/target/lun tuple, so don't print that for disappeared paths that are kept in the map in case they come back up.
print_path now prints #:#:#:# instead.
root [Wed, 4 May 2005 15:29:04 +0000 (17:29 +0200)]
[multipathd]
path_discovery() and store_pathinfo() need the "|DI_WWID" flag for set_path_owner to work().
This should fix the problem with proactive path failing not working.
root [Wed, 4 May 2005 11:03:19 +0000 (13:03 +0200)]
[multipathd]
path_discovery() should at least fetch DI_SYSFS to set the dev_t for checkers
Also removed unused sysfs_path param in path_discovery_locked() prototype
root [Wed, 4 May 2005 10:47:08 +0000 (12:47 +0200)]
[multipath]
'-l' flag now displays only info provided by sysfs and DM
'-ll' flag displays what '-l' used to display
Man page updated accordingly
root [Wed, 4 May 2005 08:38:42 +0000 (10:38 +0200)]
[multipath]
Update version string to 0.4.5
Christophe Varoqui [Tue, 3 May 2005 21:45:17 +0000 (14:45 -0700)]
[multipath][multipathd]
Remove remnents of the '-S' flag handling.
Benjamin Marzinski, RedHat
Christophe Varoqui [Tue, 3 May 2005 21:18:55 +0000 (14:18 -0700)]
[multipathd]
Apply blacklist to uevents too.
Indeed, we don't want too add a checker on /dev/dm-0 when this devmap comes up.
Christophe Varoqui [Mon, 2 May 2005 21:19:48 +0000 (14:19 -0700)]
[multipathd]
Log only once the failure messages in loops.
Should close https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=156641
Christophe Varoqui [Mon, 2 May 2005 21:05:34 +0000 (14:05 -0700)]
[multipathd]
Definitely kill the "initial map reconfigure" at daemon startup.
The daemon shouldn't do surprising things like that.
Christophe Varoqui [Mon, 2 May 2005 20:55:57 +0000 (13:55 -0700)]
[libmultipath]
Clariion arrays need the "queue_if_no_path" feature for the system to survive a firmware rolling upgrade. Set it by default in hwtable.c
Suggested by Edward Goggin, EMC
Christophe Varoqui [Sun, 1 May 2005 23:17:46 +0000 (16:17 -0700)]
[multipath]
Check for root privilege before running.
Avoids failing on first DM ioctl.
Christophe Varoqui [Sun, 1 May 2005 22:10:03 +0000 (15:10 -0700)]
[build]
Don't include ".git" in recursive make
Christophe Varoqui [Sun, 1 May 2005 22:05:22 +0000 (15:05 -0700)]
Initial git import.
Release 0.4.5-pre2