Jason Dillaman

Restrict the number of LUNs to 256

2

This is the maximum supported by Linux and defining more than 256 images results in the rbd-target-gw crashing and burning

... it should run under ```ceph``` user

Signed-off-by: Jason Dillaman

With a clean configuration, the daemon fails with the following error: > Traceback (most recent call last): > File "/bin/rbd-target-api", line 783, in > config = Config(logger) > File "/usr/lib/python2.7/site-packages/ceph_iscsi_config/common.py",...

The default settings specify a fully qualified path for the keyring [1] which is concatenated with "/etc/ceph" when passing the keyring path to Rados [2]. [1] https://github.com/pcuzner/ceph-iscsi-config/blob/master/ceph_iscsi_config/settings.py#L19 [2] https://github.com/pcuzner/ceph-iscsi-config/blob/master/ceph_iscsi_config/common.py#L35

rbd-target-api daemon runs as privileged user

2

... it should run under ```ceph``` user

Prometheus dashboards are tied to "Local" datasource

4

The ansible installer configures a "prometheus" datasource, so lots of dashboards fail to load due to the missing data. These should either just use the "default" datasource or be switched...

Network utilization filters out standard 10GbE device names

2

... a la p1p1, p5p2, etc.

Per-OSD dashboard should include read and write latency

3

Something like the following: `irate(ceph_osd_op_r_latency_sum[1m]) / on (ceph_daemon) irate(ceph_osd_op_r_latency_count[1m]) * 1000 ` `irate(ceph_osd_op_w_latency_sum[1m]) / on (ceph_daemon) irate(ceph_osd_op_w_latency_count[1m]) * 1000 ` It might also be nice to have a global latency...