Error returned from deploy ramdisk: Could+not+find+disk+to+use

[Naohiro Tamura]

pxe_ipmitool iscsi deploy failed due to "Error returned from deploy ramdisk: Could+not+find+disk+to+use".

The root cause of this problem is that the bare metal server has its local hard disk as /dev/sdc.
Please find the attached image.
https://launchpadlibrarian.net/198238406/deploy_ramdisk_error_sdc.png

2015-02-20 10:51:29.563 DEBUG oslo_messaging._drivers.amqp [-] UNIQUE_ID is da7fcb58dd6a4ba4af8a02c59665fb8c. from (pid=15228) _add_unique_id /usr/local/lib/python2.7/dist-packages/oslo_messaging/_drivers/amqp.py:224
2015-02-20 10:51:29.565 ERROR ironic.drivers.modules.iscsi_deploy [-] Error returned from deploy ramdisk: Could+not+find+disk+to+u se.
2015-02-20 10:51:29.565 DEBUG ironic.common.states [-] Exiting old state 'deploying' in response to event 'fail' fr om (pid=15228) on_exit /opt/stack/ironic/ironic/common/states.py:157

Tentative fix can be as below:

diff --git a/ironic/drivers/modules/iscsi_deploy.py b/ironic/drivers/modules/iscsi_deploy.py
index d929f5a..98a6309 100644
--- a/ironic/drivers/modules/iscsi_deploy.py
+++ b/ironic/drivers/modules/iscsi_deploy.py
@@ -66,7 +66,7 @@ pxe_opts = [
                help='Maximum TTL (in minutes) for old master images in '
                'cache.'),
     cfg.StrOpt('disk_devices',
- default='cciss/c0d0,sda,hda,vda',
+ default='cciss/c0d0,sda,sdc,hda,vda',
                help='The disk devices to scan while doing the deploy.'),
     ]

[Naohiro Tamura (naohirot)]

If fundamental fix needs certain amount of time and effort, please assign this to me, I'll submit the above tentative fix as a short term solution.

Thanks.

Naohiro

[Lucas Alvares Gomes (lucasagomes)]

Adding TripleO as one of the projects affected because we will need to fix it in the find_disk() function of the DIB tool.

Right now the DIB tool look at the disk= kernel parameter which is passed by Ironic to find the disk, and it relies on the disk names. I think that DIB should just try to find a suitable disk, and not rely on any names.

It's important to note that Ironic will offer a way to pass hints to the deploy ramdisk in order to make it pick the right disk (if required), and that's explainded in more details in the spec merged in Ironic which is here http://specs.openstack.org/openstack/ironic-specs/specs/kilo/root-device-hints.html

[Naohiro Tamura (naohirot)]

Lucas,

Thank you for the comment and the information about Root device hints.

I read the Root device hints bp, source, and test cases.
And I understood that Root device hints improves the case that multiple boot devices are found, but not the case that no boot device is found.

Naohiro

[Naohiro Tamura (naohirot)]

Lucus,

Would you set the "Importance" of this bug report?

I'd like to ask you to set "High", because FUJITSU PRIMERGY server cannot be deployed.

Thanks.

Naohiro

[Ramakrishnan G (rameshg87) (rameshg87)]

Naohiro, if it's just adding sdc to the list of 'disk_devices', then you can actually set it in your environment in /etc/ironic/ironic.conf and then restart conductor.

I think it's still useful if the operator can specifyt the disk_devices because linux kernel might name the block devices based on it's type - one some systems it might /dev/sdX, on some other /dev/vdX, on some other /dev/cciss.

We should enhance diskimage-builder to support glob options in disk_device. Something like below:
disk_device='cciss/c[0-9]d[0-9],sd[a-z],hd[a-z],vd[a-z]'

[Ruby Loo (rloo)]

I'm going to lower this to medium because high is described as ' High if the bug prevents a key feature from working properly for some users (or with a workaround)' [1], and the work around is to change the value of the config option (as mentioned by Ramakrishnan).

[1] https://wiki.openstack.org/wiki/BugTriage

[Jim Rollenhagen (jim-rollenhagen)]

Since we're deprecating the bash ramdisk, and IPA will handle this case fine... I'm going to just go ahead and close this out.

[Jim Rollenhagen (jim-rollenhagen)]

Feel free to reopen if you disagree :)