Merge lp:~strikov-deactivatedaccount/curtin/multipath-2 into lp:~curtin-dev/curtin/trunk
Proposed by
Scott Moser
Status: | Merged |
---|---|
Merged at revision: | 211 |
Proposed branch: | lp:~strikov-deactivatedaccount/curtin/multipath-2 |
Merge into: | lp:~curtin-dev/curtin/trunk |
Diff against target: |
141 lines (+95/-0) 3 files modified
curtin/block/__init__.py (+53/-0) curtin/commands/block_meta.py (+5/-0) curtin/commands/curthooks.py (+37/-0) |
To merge this branch: | bzr merge lp:~strikov-deactivatedaccount/curtin/multipath-2 |
Related bugs: |
Reviewer | Review Type | Date Requested | Status |
---|---|---|---|
curtin developers | Pending | ||
Review via email: mp+261121@code.launchpad.net |
To post a comment you must log in.
So, some thoughts here.
a.) we're using 'strip' on the output of /lib/udev/scsi_id which is less than ideal. that was done to drop the trailing '\n'. the result is that we'd recognize ' DISK 1 ' and 'DISK 1' as the same device and take the "enable_multipath" path. Both not very likely and not very big deal, but odd. I guess probably we should:
if len(scsi_id) and scsi_id[-1] == '\n':
scsi_id = scsi_id[0:-1]
not a big deal, but worth fixing
b.) the other hting i'm concerned about is if the image / environment that we're installing in has multipath-tools installed, it may see that /dev/sda (and /dev/sde) is used by /dev/dm-0. and then choose not to install to it, or not realize that and fail to install.
in that scenario, i think the first install possibly works, while subsequent ones fail.
we may need a way to un-setup the dm devices so we can install to the "raw" ones.