820825 - please reimage 66 pandas (6 chassis) with raw android image

Status: RESOLVED FIXED

(Infrastructure & Operations :: DCOps, task)

Description

[Joel Maher ( :jmaher ) (UTC -8)]

We need to get testing and the next step is to get this latest/final image of android on some panda boards.  This is not using lifeguard, but the old fashioned method of a raw flash of the .img file.

The image is located here:
http://people.mozilla.org/~jmaher/panda/panda_2gb_v2.img.bz2

I have validated this in a series of pandas in chassis 2.

Comment 1

[Van Le [:van]]

:jmaher, is there any specific 6 chassis we should install this new image in?

Comment 2

[Derek Moore [:dmoore]]

Also, we can arrange to manually image 6 chassis for a special case like this, but it's simply not sustainable in the long term. I'm comfortable with this request, as long as it isn't the final production solution.

Comment 3

[Joel Maher ( :jmaher ) (UTC -8)]

no specific ones that I know of.  We should avoid chassis 2 and 3.  I am not sure what other chassis are used for b2g and cedar android testing.  My take is use whatever 6 are the most accessible.

:kmoir or :dividehex, any other information we should consider?

Comment 4

[Joel Maher ( :jmaher ) (UTC -8)]

this is the same solution we have used for the tegras over the last 3 years.  We are working on a solution similar to what we are doing for b2g, it isn't ready.  For the near future I see us doing a lot of hand flashing while we ramp up the panda android testing and sometime next quarter we will have a lifeguard based solution for android.

Comment 5

[Justin Wood (:Callek)]

(In reply to Joel Maher (:jmaher) from comment #3)
> no specific ones that I know of.  We should avoid chassis 2 and 3.  I am not
> sure what other chassis are used for b2g and cedar android testing.  My take
> is use whatever 6 are the most accessible.
> 
> :kmoir or :dividehex, any other information we should consider?

My opinion is we should make sure that at *least* half are not-in-same-rack/vlan. Such that if a single switch goes down we don't lose the whole pool.

:arr since I know there are still some issues with individual relay boards, and that you're also not done configuring relay boards, are there specific chasis's you can recommend?

Comment 6

[Kim Moir [:kmoir] ET]

Regarding comment 3, here is the current state

chassis 2 (panda-{0022..0033}): releng staging + joel 
chassis 3 (panda-{0034..0045}): relops
chassis 4 (panda-{0046..0057}): kmoir (cedar tests in prod) 
chassis 5 (panda-{0058..0069}): kmoir (cedar tests in prod)
Chassis-0[07-14] and panda-00[82-169] b2g android

Comment 7

[Van Le [:van]]

The image has been deployed to Pandas-[0808-0873] which are panda-chassis-[073-078].

Comment 8

[Amy Rich [:arr] [:arich]]

Note that you *MUST* set the state of pandas to locked_out when you hard code an image on them.  Otherwise mozpool will still happily try to control them and reboot them and attempt to reimage them at will.  I have set panda-{0808..0873} in the locked out state and added a comment about this bug for now.

Comment 9

[Joel Maher ( :jmaher ) (UTC -8)]

Are these using the hardcoded image file (i.e. dd if=android.img of=/dev/sdb) instead of the preseed image which we use for b2g and lifeguard?

Comment 10

[Van Le [:van]]

these are using the image you linked in comment 1.

>The image is located here:
>http://people.mozilla.org/~jmaher/panda/panda_2gb_v2.img.bz2

Van