Re: AOE/vblade session shutdown
On Wed, May 16, 2012 at 07:36:32PM -0400, Simon Kirby wrote:
>On a semi-related note,
>
>I've always wondered why AOE doesn't have an un-advertisment for offlined
>luns. Onlining a lun sends out a packet, so why not the reverse.
The protocol seems to include support for this already:
http://support.coraid.com/documents/AoEr11.txt
Section 2.4 ("Error") reads:
If Flags bit E is set, this field contains an error code defined as
follows:
[...]
Error 3: Device unavailable
The server can no longer accept ATA commands.
[...]
>If any AOE clients have the device open, they can just ignore it like
>they currently do anyway (and wait for aoe_deadsecs before kicking up the
>failure). The rest could just flush the device and pretend it never
>existed.
>
>The current behaviour can be dangerous because certain other typical
>tools can stumble over an old lun unexpectly at some later time. For
>example, running of "fdisk -l", or UUID scanning when doing HA fail-over,
>etc. There are lots of ways to trip on this. We have to also be careful
>when making upate luns to go and flush them from all hosts so that they
>don't hang later when initiating a fail-over.
>
>I suppose for completeness, the same packet should be sent as a shelf
>reboots (or vblade exits), which should be handled the same way by the
>clients. This would take care of left-over update luns, too.
>
>Simon-
>
>On Tue, May 15, 2012 at 06:57:21PM +0200, Alexandre wrote:
>
>> I usually use aoe-flush to stop using an aoe block device on a client.
>> Doing so has the great advantage of deleting an exported device from a
>> client and still use others from other target (or not). While unloading the
>> module shuts everything down.
>> Flushing aoe devices require recent aoe module, won't work on stock Kernel
>> AFAIR.
>> Le 15 mai 2012 17:38, "Charles Chou" <cchou@...> a ?crit :
>>
>> > Hi,
>> > What is the proper way to gracefully shutdown an AOE session? I
>> > have been just killing the vblade processes on the server and rmmod aoe
>> > on the client but this seems unnatural.
>> >
>> > Also vblade is run as a foreground process as default. To make it
>> > run as a daemon I had to just put it in background with an & (I need to
>> > export more than one target). Is this the right way?
>> >
>> > Lastly, where does the vblade log go? Do I need to explicitly do a
>> > re-direct of stdout/stderr?
>> >
>> > Thanks,
>> > Charles Chou
>> >
>> >
>> > ------------------------------------------------------------------------------
>> > Live Security Virtual Conference
>> > Exclusive live event will cover all the ways today's security and
>> > threat landscape has changed and how IT managers can respond. Discussions
>> > will include endpoint security, mobile security and the latest in malware
>> > threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
>> > _______________________________________________
>> > Aoetools-discuss mailing list
>> > Aoetools-discuss@...
>> > https://lists.sourceforge.net/lists/listinfo/aoetools-discuss
>
>------------------------------------------------------------------------------
>Live Security Virtual Conference
>Exclusive live event will cover all the ways today's security and
>threat landscape has changed and how IT managers can respond. Discussions
>will include endpoint security, mobile security and the latest in malware
>threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
>_______________________________________________
>Aoetools-discuss mailing list
>Aoetools-discuss@...
>https://lists.sourceforge.net/lists/listinfo/aoetools-discuss
--
--
Jesse Becker
NHGRI Linux support (Digicon Contractor)
:(){ :&:};:
------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
RSS Feed