can't setup vstart cluster on Fedora 33

20 Apr 2021

      Hi,

I am running Fedora 33 and am failing to get vstart cluster running on it.
Running "../src/vstart.sh -n -d" ends up in an infinite loop and following
two lines get printed on stdout at every iteration -

/home/rishabh/repos/ceph/test2/build/bin/ceph -c
/home/rishabh/repos/ceph/test2/build/ceph.conf -k
/home/rishabh/repos/ceph/test2/build/keyring -h
waiting for mgr restful module to start

I've seen vstart.sh loop at the exact same point a few months ago when few
Python dependencies were missing.

I took a look into build/out/mgr.x.log and the log message
there seems to be speak of the cause -
https://gist.github.com/rishabh-d-dave/c4d413fcdf67aadffc2875479cd5d73f.
Since the log message includes the names of the MGR plugins and not some
Python packages/modules, I think the issue is most probably a bug somewhere
in Ceph repo. Here are the few more related messages from the same log during
same run - https://gist.github.com/rishabh-d-dave/05c75bb4729014295f32706a0555b17e

I've tried running vstart.sh differently -- passing the number of daemons
explicitly, running with different options, etc -- but nothing has led to a
different outcome so far. I've also tried twice to recompile the code from
scratch -- triggering building after "git clean" andsubmodule update and
building on a brand new repo once. I was at the latest master orvery close to
it both the times. The commit IDs are
477b705fe782668427bd11ab3fe9e350b9e72eac and
0be78da368f2dc1c891e3caafac38f7aa96d3c49.

I used following commands to trigger code's compilation -

time(sudo ./install-deps.sh &&
./do_cmake.sh -DWITH_CEPHFS_SHELL=ON -DWITH_BABELTRACE=OFF
-DWITH_MANPAGE=OFF -DWITH_RBD=OFF -DWITH_RADOSGW=OFF -DWITH_KRBD=OFF
&&
cd build &&
ninja -j 7
)

Is this issue already known/seen? I haven't opened a ticket on the tracker
since I am not sure that this is definitely a bug.

Thanks,
- Rishabh

Rishabh Dave

jlayton＠redhat.com

tags

participants (2)