It was trying to pass multiple paths to fusermount, but
it should instead invoke fusermount for each path.
Signed-off-by: John Spray <john.spray@redhat.com>
If you do something like this:
- foo:
- bar:
- baz
Then catching AttributeError is necessary for giving a somewhat
friendlier error on the 'baz' where we have a string instead
of a dict.
Signed-off-by: John Spray <john.spray@redhat.com>
The hardcoded arguments are deprecated for the profile name. Create a
profile that matches the previous default arguments. Read the config to
allow the caller to override the defaults from the yaml file.
Signed-off-by: Loic Dachary <loic@dachary.org>
The former lets you specify a specific ceph-qa-suite branch to use for
testing. The latter lets you specify an as-is directory to use, for
example if you want ro run tests that you don't want to commit yet.
Signed-off-by: Zack Cerza <zack.cerza@inktank.com>
When used in the radosgw context, the 'ceph' string must be present but
when used in ceph_manager, the 'ceph' string is already included.
Signed-off-by: Loic Dachary <loic@dachary.org>
if the value from the yaml file is an int, it must be converted into a
string bevore being concatenated with '='
Signed-off-by: Loic Dachary <loic@dachary.org>
Instead of being hard coded, the erasure code profile is read from yaml
as:
erasure_code_profile:
name: teuthologyprofile
k: 2
m: 1
ruleset-failure-domain: osd
Which translates into
ceph osd erasure-code-profile set teuthologyprofile \
k=2 m=1 ruleset-failure-domain=osd
The semantic of k/m etc. depends on the erasure code plugin, but it is
common to use k as the dividing factor for each object and m as the
number of coding chunks.
Using a different set of parameters or a different plugin can
be done without modifying the teuthology code:
erasure_code_profile:
name: LRCprofile
plugin: LRC
mapping: __DD__DD
layers: [[ "_cDD_cDD", "" ], [ "cDDD____", "" ], [ "____cDDD", ""]]
For backward compatibility, the default erasure code profile is set to
k: 2
m: 1
ruleset-failure-domain: osd
which matches the previously hardcoded default of the corresponding
functions.
Signed-off-by: Loic Dachary <loic@dachary.org>
This is a contextmanager task that will run some command in the background
for the duration of any subsequent tasks, and kill it in the cleanup
phase.
Signed-off-by: Sage Weil <sage@inktank.com>
New ceph versions will no longer create the
CephFS pools and will not enable the filesystem
by default. This change adds an explicit step
to the cluster initialization to create the
filesystem if it doesn't already exist.
Signed-off-by: John Spray <john.spray@inktank.com>
