* Itamar Heim <iheim(a)redhat.com> [2012-09-27 02:21]:
On 09/24/2012 02:46 PM, Ryan Harper wrote:
>* Sheldon <shaohef(a)linux.vnet.ibm.com> [2012-09-24 06:50]:
>>On 09/21/2012 08:14 PM, Laszlo Hornyak wrote:
>>>
>>>----- Original Message -----
>>>>From: "Sheldon" <shaohef(a)linux.vnet.ibm.com>
>>>>To: vdsm-devel(a)lists.fedorahosted.org
>>>>Sent: Friday, September 21, 2012 11:42:05 AM
>>>>Subject: [vdsm] how engine get files from node???
>>>>
>>>>
>>>>Hi all,
>>>>I have submitted a patch about watchdog device
>>>>http://gerrit.ovirt.org/#/c/7535/
>>>>If we set 'dump' action for watchdog, qemu will generate a dump
file.
>>>>And I have get some feedback, how engine get these files?
>>>When adding the host (if installation is enabled) the engine does some
ssh/scp to the new host, but this can be turned off. Also, the guideline is to have a
single interface with the host, and after installation the engine does not try to
communicate with vdsm in any other way than the xmlrpc interface.
>>
>>do you means "ssh/scp" may not be available on host, So a xmlrpc
>>interface should be added to get the files on host?
>
>ssh/scp is available, but requires user authentication. We already have
>one entry point into the node, the VDSM API, which is currently running
>over XMLRPC, so, instead of attempting to coordinate additional users
>and authenication, provide a way to programatically get at the data via
>XMLRPC.
is this needed by the engine interactively, or just for log colleciton?
logs/cores/dumps are collected today by the log collector utility,
based on ssh/scp leveraging sos.
but this isn't "engine getting logs", rather a stand alone utility
for collecting them.
is the file needed by engine for some flow, or just log collection
for user benefit?
I can imagine both scenarios. In general, if the node has
cores/logs/dumps for a particular VM, it would be nice to have those
available for debugging, reporting issues, etc. We always have
scenarios where we wish we had a better picture of what was happening on
the system at the time of failure. For vdsm, we do since we have host
logs/stats as well as anything we captured from the guest (cores, etc).
I'd like a way for the mgmt server to specify either, all of the host
logs, or a specific guest log, and fetch all of that data from the
end-point. Certainly something with ssh/scp can be worked out, but it
would be nice to have this available programatically.
Think of the engine flow around a VM that's tagged as HA. If engine
detects a failure, it will create the VM again on a different host; then
engine could pull debugging data and log files from the host to store
that for the admin to push to her support connection.
And instead of having the admin know to run this stand alone utility, it
could be tied into an existing HA recover use-case.
Another case some of our mgmt products use is to do collect error
reports and statistics immediately after a failure has occured and
stores them on the management server, raising a notice to the admin for
submission to support.
These use-cases are easier to drive with an API call rather than the
standalone utility.
Now, on the contrary; if vdsm itself is having issues, then it seems
that such a separate utility is still needed.
>
>>
>>>>scp? or Will vdsm support new API to list and get these files?
>>>>IMO, The new API should list and get not only dump files, but also
>>>>other kinds of files.
>>>>
>>>>
>>>>--
>>>>Sheldon Feng <shaohef(a)linux.vnet.ibm.com> IBM Linux Technology
Center
>>>>_______________________________________________
>>>>vdsm-devel mailing list
>>>>vdsm-devel(a)lists.fedorahosted.org
>>>>https://lists.fedorahosted.org/mailman/listinfo/vdsm-devel
>>>>
>>
>>
>>--
>>Sheldon Feng(?????????)<shaohef(a)linux.vnet.ibm.com>
>>IBM Linux Technology Center
>
--
Ryan Harper
Software Engineer; Linux Technology Center
IBM Corp., Austin, Tx
ryanh(a)us.ibm.com