Closed Bug 811316 Opened 12 years ago Closed 12 years ago

update panda live boot squashfs image

Categories

(Infrastructure & Operations :: RelOps: General, task)

task
Not set
normal

Tracking

(Not tracked)

RESOLVED FIXED

People

(Reporter: dustin, Assigned: dividehex)

References

Details

Attachments

(1 file)

I've found a few problems so far: * mobile-init.sh sleeps for 30s when it begins. There's no comment to indicate why this would be useful. Can we remove it? * It would be great to send a lifeguard event when mobile-init.sh starts * More worryingly, the rsyslog config file uses ntp.build.mozilla.com, rather than ntp.build.mozilla.org, which causes rsyslog to not start, which causes mobile-init to not start.
*slight* correction to point 3: the bogus hostname does *not* cause rsyslog to fail. Omitting syslog-server= *does* - my error.
Another fix: * use PST timezone, at least as long as everything else is running in PST
Two more: * panic() doesn't quote the log argument correctly, so instead of "second-stage.sh returned non-zero", you just get Nov 13 16:45:52 panda-0045 mobile-init.sh: second-stage.sh * it'd be great if this would log the error output of second-stage.sh in this case
These are all baked into the live boot squashfs file, not the preseed image. I'll also add some documentation as to unpacking/editing/repacking the squashfs.
Summary: update panda preseed image → update panda live boot squashfs image
Oh, silly me. And bonus, because that's easier to change!
* We'll also need a full Python
* We'll also need uboot-mkimage or (perhaps better) u-boot-tools
Blocks: 812251
New sqaushfs has been compiled with the following changes: * changed timezone to US/Pacific * install python2.7 * install curl * install u-boot-tools (mkimage) * changed ntp.build.mozilla.org in pre-syslog.sh * mobile-init.sh: properly quotes argument * mobile-init.sh: logs second-stage.sh err output on non 0 exit * mobile-init.sh: removed 30 sleep Changes still needed: * Lifeguard event needed for mobile-init.sh I'm at an airport and just didn't have time to add this last one in. I can add it in on the second pass.
Looks good! I've installed this (with the old image in /tmp on releng-puppet1.srv.releng.scl3), and will submit a patch to take out my workarounds in bug 812251.
I just copied this over the existing squashfs, but it occurs to me that we should be pointing to the datestamped version using the PXE configs, so that we can test and revert if necessary. Next time!
per dustin via irc: the Lifeguard event will be "mobile_init_started"
Some tweaks to mobile-init.sh in bug 816732, too
mkdosfs needs to be installed. bug 817832
Adds lifeguard event mobile_init_started to mobile-init.sh
Attachment #689008 - Flags: review?(dustin)
FTR: I've uploaded a new squashfs ( precise-panda-live-build.2012120401.squashfs ) with the double ntpdate calls fixed (bug 816732) and mkdosfs installed. This DOES NOT include the mobile-init.sh patch in c#14.
Comment on attachment 689008 [details] [diff] [review] lifeguard event call added I like this. The code duplication is annoying, but necessary in this case.
Attachment #689008 - Flags: review?(dustin) → review+
New squashfs is uploaded: precise-panda-live-build.2012120602.squashfs * updated cltbld passwd * added lifeguard event "mobile_init_started" * mobile-init.sh waits for ntpdate to adjust time before starting
All active pxe configs have been updated to the new squashfs.
Status: NEW → RESOLVED
Closed: 12 years ago
Resolution: --- → FIXED
Component: Server Operations: RelEng → RelOps
Product: mozilla.org → Infrastructure & Operations
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: