Yeah, I noted I was looking at the wrong lockspace. The gfs2 lockspace in
this cluster is vol1. Once I corrected at what I was looking at, I think I
solved my problem: I believe the problem is an endian thing. In
set_rcom_status:
rs->rs_flags = cpu_to_le32(flags)
However, in receive_rcom_status() flags are checked:
if (!(rs->rs_flags & DLM_RSF_NEED_SLOTS)) {
But it should be:
if (!(le32_to_cpu(rs->rs_flags) & DLM_RSF_NEED_SLOTS)) {
I made this change and now the gfs2 volume is being mounted correctly on
both nodes. I¹ve repeated it a number of times and it¹s kept working.
Neale
On 10/14/14, 3:20 PM, "David Teigland" <teigland@xxxxxxxxxx> wrote:
>clvmd is a userland lockspace and does not use lockspace_ops or slots/jids
>like a gfs2 (kernel) lockspace.
>
>To debug the dlm/gfs2 control mechanism, which assigns gfs2 a jid based on
>dlm slots, enable the fs_info() lines in gfs2/lock_dlm.c. (Make sure that
>you're not somehow running gfs_controld on these nodes; we quit using that
>in RHEL7.)
--
Linux-cluster mailing list
Linux-cluster@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/linux-cluster
