DanielSoares
Contributor
Contributor

what happens when loosing the storage with ESX4

Hello

I had a very interresting issue yesterday

the san controlers rebooted so we loosed access to all the luns during 4 minutes

result big issues on all serveurs using san storage except all the virtual machines ...

the esx logs tells that they loosed the storage connection and immediatly the virtual machines went to suspend state

the result of this was that the vms continued to work perfectly after the crash

this was confirmed by the vmware support but they have no more informations

can some one explain exactly the mechanism of this process? are they documentations about this very interesting solution?

is there a timeout after what the vm crashes ?

thanks for everybody who have infos

0 Kudos
7 Replies
AWo
Immortal
Immortal

Are you sure that the controllers rebooted at the same time? Are the other hosts which were facing problem dual-attached and do they have path-failover software running (like PowerPath or SecurePath, or mpio)?


AWo

VCP 3 & 4

\[:o]===\[o:]

=Would you like to have this posting as a ringtone on your cell phone?=

=Send "Posting" to 911 for only $999999,99!=

vExpert 2009/10/11 [:o]===[o:] [: ]o=o[ :] = Save forests! rent firewood! =
0 Kudos
mcowger
Immortal
Immortal

Agreed - are you sure both controllers rebooted at the same time?

If so, your guest OSs basically saw long delays in accessing their storage, and if they are configured properly, may not have crashed. depends on config.

--Matt

VCP, VCDX #52, Unix Geek, Storage Nerd

9773_9773.gif

--Matt VCDX #52 blog.cowger.us
0 Kudos
DanielSoares
Contributor
Contributor

both controlers rebooted at same time yes

all luns where loosed during 4 minutes

and no issues on the VMs ... the ESXs puted them on suspend mode until the storage beeing available again

0 Kudos
mcowger
Immortal
Immortal

ESX actually doesn't put them in 'suspend' mode. It just held all IO, and your guest OSs handled it well Smiley Happy

--Matt

VCP, VCDX #52, Unix Geek, Storage Nerd

9773_9773.gif

--Matt VCDX #52 blog.cowger.us
0 Kudos
DanielSoares
Contributor
Contributor

do you mean that the guest continue to work in term of cpu and memory?

how can they continue without IO ?

what happen if one process need access to a file?

0 Kudos
mcowger
Immortal
Immortal

do you mean that the guest continue to work in term of cpu and memory?

YES

how can they continue without IO ?

They can continue if they dont need IO

what happen if one process need access to a file?

That varies by OS. Linux would likely reject the IO eventually and throw the filesystem into read only mode. Windows can often blue screen, but it all depends on settings.

--Matt

VCP, VCDX #52, Unix Geek, Storage Nerd

9773_9773.gif

--Matt VCDX #52 blog.cowger.us
0 Kudos
AWo
Immortal
Immortal

IMHO you just had luck.


AWo

VCP 3 & 4

\[:o]===\[o:]

=Would you like to have this posting as a ringtone on your cell phone?=

=Send "Posting" to 911 for only $999999,99!=

vExpert 2009/10/11 [:o]===[o:] [: ]o=o[ :] = Save forests! rent firewood! =
0 Kudos