clusterhq / ceph-flocker-driver Goto Github PK

Ceph supplies two Python libraries: rbd.py and rados.py. On Ubuntu, these are installed into /usr/lib/python2.7/dist-packages and provided by the packages python-rados and python-rbd.

Flocker operated in a virtualenv at /opt/flocker. This prevents Flocker from accessing these libraries. Unfortunately this driver requires them.

Currently, I am using flocker ceph and mesos together. And I use flocker and ceph as the storage solution for my mesos cluster.

But it seems that flocker does not work stable enough. The command

docker run --volume-flocker=ceph -v xxx:/xxx some-image

may get blocked and show message Timed out waiting for dataset to mount. I checked the log of flocker and I still don't know the reason. The json format log is really hard to understand.

I have multi flocker agent running. And only one of them failed to run container with flocker as volume driver. Others work fine.

As per the title - I'm really interested in seeing this driver make it to release. How's it going and is it likely to be out of WIP soon?

is there any people still develop it？

I see lots of errors in node logs for rbd showmapped failing. Looks like the syntax is wrong.

ceph_rbd.py, line 121;
[b"rbd", "-p", self._pool, b"showmapped"]).strip()

root@lumere-controller-1:~# rbd -p rbd showmapped
rbd: unrecognised option '-p'

I think line 121 should look like;
[b"rbd", b"showmapped"]).strip()

I even don't know the backend name when use ceph with flocker

All interactions with RBD should be through something like a a MappingRBD interface, with a verified fake. This will allow us to unit and integration test the entire driver.

It seems that this storage driver currently only works on ceph-infernalis and not ceph-jewel.

I can have a look and try and clean up and fix the problem.

I have forked the repo and will try and fix it during the weekend :)

Does not work

acehege@docker001:~$ ceph --version
ceph version 10.2.2 (45107e21c568dd033c2f0a3107dec8f0b0e58374)

acehege@lspdocker001:~$ sudo cat /etc/flocker/agent.yml
"version": 1
"control-service":
   "hostname": "docker001"

"dataset":
    "backend": "ceph_flocker_driver"
    "user_id": "ceph"
    "cluster_name": "ceph"
    "ceph_conf_path": "/etc/ceph/ceph.conf"

Works

acehege@docker002:~$ ceph --version
ceph version 9.2.1 (752b6a3020c3de74e07d2a8b4c5e48dab5a6b6fd)

acehege@docker002:~$ sudo cat /etc/flocker/agent.yml
"version": 1
"control-service":
   "hostname": "docker001"

"dataset":
    "backend": "ceph_flocker_driver"
    "user_id": "ceph"
    "cluster_name": "ceph"
    "ceph_conf_path": "/etc/ceph/ceph.conf"