Month: March 2015

When a teuthology target (i.e. machine) is provisioned with teuthology-lock for the purpose of testing Ceph, there is no way to choose the kernel. But it can be installed afterwards using the following:

cat > kernel.yaml <<EOF
interactive-on-error: true
roles:
- - mon.a
  - client.0
kernel:
   branch: testing
tasks:
- interactive:
EOF

Assuming the target on which the new kernel is to be installed is vpm083, running

$ teuthology  --owner loic@dachary.org \
  kernel.yaml <(teuthology-lock --list-targets vpm083)
...
2015-03-09 17:47 INFO:teuthology.task.internal:Starting timer...
2015-03-09 17:47 INFO:teuthology.run_tasks:Running task interactive...
Ceph test interactive mode, use ctx to interact with the cluster
>>>

will install an alternate kernel and reboot the machine:

[ubuntu@vpm083 ~]$ uname -a
Linux vpm083 3.19.0-ceph-00029-gaf5b96e #1 SMP Thu Mar 5 01:04:25 GNU/Linux
[ubuntu@vpm083 ~]$ lsb_release -a
LSB Version:	:base-4.0-amd64:base-4.0-noarch:
Distributor ID:	RedHatEnterpriseServer
Description:  release 6.5 (Santiago)
Release:	6.5
Codename:	Santiago

Command line arguments to the kernel may be added to /boot/grub/grub.conf. For instance loop.max_part=16 to allow partition creation on /dev/loop devices:

default=0
timeout=5
splashimage=(hd0,0)/boot/grub/splash.xpm.gz
hiddenmenu
title rhel-6.5-cloudinit (3.19.0-ceph-00029-gaf5b96e)
        root (hd0,0)
        kernel /boot/vmlinuz-3.19.0 ro root=LABEL=79d3d2d4  loop.max_part=16
        initrd /boot/initramfs-3.19.0.img

When handling a Ceph OSD, it is convenient to assign it a symbolic name that can be chosen even before it is created. That’s what the uuid argument for ceph osd create is for. Without a uuid argument, a random uuid will be assigned to the OSD and can be used later. Since the ceph osd create uuid is idempotent, it can also be used to lookup the id of a given OSD.

$ osd_uuid=b2e780fc-ec82-4a91-a29d-20cd9159e5f6
# convert the OSD uuid into an OSD id
$ ceph osd create $osd_uuid
0
# convert the OSD id into an OSD uuid
$ ./ceph --format json osd dump | jq '.osds[] | select(.osd==0) | .uuid'
"b2e780fc-ec82-4a91-a29d-20cd9159e5f6"

The Ceph integration tests may fail because of environmental problems (network not available, packages not built, etc.). If six jobs failed out of seventy, these failed test can be re-run instead of re-scheduling the whole suite. It can be done using the **–filter** option of teuthology-suite with a comma separated list of the job description that failed.
The job description can either be copy/pasted from the web interface or extracted from the paddles json output with:

$ run=loic-2015-03-03_12:46:38-rgw-firefly-backports---basic-multi
$ paddles=http://paddles.front.sepia.ceph.com
$ eval filter=$(curl --silent $paddles/runs/$run/jobs/?status=fail |
  jq '.[].description' | \
  while read description ; do echo -n $description, ; done | \
  sed -e 's/,$//')

Where the paddles URL outputs a json description of each job of the form:

[
  {
    "os_type": "ubuntu",
    "nuke_on_error": true,
    "status": "pass",
    "failure_reason": null,
    "success": true,
...
    "description": "rgw/multifs/{clusters/fixed-2.yaml}"
  },
  {
    "os_type": "ubuntu",
...

The jobs/?status=fail part of the URL selects the jobs with “success”:false. The jq expression displays the description field (.[].description), one by line. These lines are aggregated into a comma separated list (while read description ; do echo -n $description, ; done) and the trailing comma is stripped (sed -e ‘s/,$//’). The filter variable is set to the resulting line and evaled to get rid of the quotes (eval filter=$(..)).
The command used to schedule the entire suite can be re-used by adding the –filter=”$filter” argument and will only run the failed jobs.

$ ./virtualenv/bin/teuthology-suite --filter="$filter" \
  --priority 101 --suite rgw --suite-branch firefly \
  --machine-type plana,burnupi,mira \
  --distro ubuntu --email loic@dachary.org \
  --owner loic@dachary.org  \
  --ceph firefly-backports
...
Suite rgw in suites/rgw scheduled 6 jobs.
Suite rgw in suites/rgw -- 56 jobs were filtered out.