bosh agent rewrites partition mistakenly during VM reboot

[cloudfoundry]

bosh agent rewrites partition mistakenly during VM reboot

[cppforlife]

Added a story: https://www.pivotaltracker.com/story/show/129727181 to properly handle the error; however, this wouldnt fix the "sfdisk: cannot open" error. Are you sure changes for multipath disk handling are properly accounting for partitions?

[mattcui]

@cppforlife Yes, it's due to multipath, I submitted PR 93, please help review. Thank you!

/cc @maximilien @dpb587-pivotal

[mattcui]

Below is the agent log after making this change, worked as expected.

2016-09-05_12:13:05.45238 [linuxPlatform] 2016/09/05 12:13:05 DEBUG - Persistent disk size to be partitioned is: 21474836480, and error is: <nil>
2016-09-05_12:13:05.45240 [linuxPlatform] 2016/09/05 12:13:05 DEBUG - fdisk partitioner was chosen
2016-09-05_12:13:05.45240 [MountPersistentDisk] partitions passed into Partition()
2016-09-05_12:13:05.45242 [[Type: linux, SizeInBytes: 0]]
2016-09-05_12:13:05.45250 [Cmd Runner] 2016/09/05 12:13:05 DEBUG - Running command 'sfdisk -d /dev/mapper/3600a098038303453692447376330554d'
2016-09-05_12:13:05.45551 [Cmd Runner] 2016/09/05 12:13:05 DEBUG - Stdout: # partition table of /dev/mapper/3600a098038303453692447376330554d
2016-09-05_12:13:05.45552 unit: sectors
2016-09-05_12:13:05.45553
2016-09-05_12:13:05.45553 /dev/mapper/3600a098038303453692447376330554d1 : start=        1, size= 41929649, Id=83
2016-09-05_12:13:05.45554 /dev/mapper/3600a098038303453692447376330554d2 : start=        0, size=        0, Id= 0
2016-09-05_12:13:05.45554 /dev/mapper/3600a098038303453692447376330554d3 : start=        0, size=        0, Id= 0
2016-09-05_12:13:05.45555 /dev/mapper/3600a098038303453692447376330554d4 : start=        0, size=        0, Id= 0
2016-09-05_12:13:05.45555 [Cmd Runner] 2016/09/05 12:13:05 DEBUG - Stderr:
2016-09-05_12:13:05.45557 [Cmd Runner] 2016/09/05 12:13:05 DEBUG - Successful: true (0)
2016-09-05_12:13:05.45557 [getPartitions] sfdisk -d result split line: /dev/mapper/3600a098038303453692447376330554d1 : start=        1, size= 41929649, Id=83
2016-09-05_12:13:05.45562 [Cmd Runner] 2016/09/05 12:13:05 DEBUG - Running command 'sfdisk -s /dev/mapper/3600a098038303453692447376330554d-part1'
2016-09-05_12:13:05.46009 [Cmd Runner] 2016/09/05 12:13:05 DEBUG - Stdout: 20964824
2016-09-05_12:13:05.46010 [Cmd Runner] 2016/09/05 12:13:05 DEBUG - Stderr:
2016-09-05_12:13:05.46011 [Cmd Runner] 2016/09/05 12:13:05 DEBUG - Successful: true (0)
2016-09-05_12:13:05.46012 [getPartitions] partition.SizeInBytes is 21467979776
2016-09-05_12:13:05.46013 [getPartitions] sfdisk -d result split line: /dev/mapper/3600a098038303453692447376330554d2 : start=        0, size=        0, Id= 0
2016-09-05_12:13:05.46015 [getPartitions] sfdisk -d result split line: /dev/mapper/3600a098038303453692447376330554d3 : start=        0, size=        0, Id= 0
2016-09-05_12:13:05.46015 [getPartitions] sfdisk -d result split line: /dev/mapper/3600a098038303453692447376330554d4 : start=        0, size=        0, Id= 0
2016-09-05_12:13:05.46021 [Cmd Runner] 2016/09/05 12:13:05 DEBUG - Running command 'sfdisk -s /dev/mapper/3600a098038303453692447376330554d'
2016-09-05_12:13:05.46319 [Cmd Runner] 2016/09/05 12:13:05 DEBUG - Stdout: 20971520
2016-09-05_12:13:05.46320 [Cmd Runner] 2016/09/05 12:13:05 DEBUG - Stderr:
2016-09-05_12:13:05.46321 [Cmd Runner] 2016/09/05 12:13:05 DEBUG - Successful: true (0)
2016-09-05_12:13:05.46324 [SfdiskPartitioner] 2016/09/05 12:13:05 INFO - /dev/mapper/3600a098038303453692447376330554d already partitioned as expected, skipping

[cppforlife]

@mattcui I believe this issue was resolved after fixing multipath disk lookup? Can be closed?

[mattcui]

@cppforlife It's addressed by my PR 93, the commit has been already merged into master branch 1.5 months ago, but it still hasn't been picked by BOSH release code, so we are still waiting for the new BOSH release which includes this fix. This fix is very important to our CF deployments on Softlayer. Thanks.

/cc @maximilien

[mattcui]

I saw the fix has already been merged in 3263.8 stemcell, closing this issue. Thanks.