On 20/03/2019 07:43, li jerry wrote:
Hi,ALL
I’ve deployed mimic(13.2.5) cluster on 3 CentOS 7.6 servers, then
configured iscsi-target and created a LUN, referring to
http://docs.ceph.com/docs/mimic/rbd/iscsi-target-cli/.
I have another server which is CentOS 7.4, configured and mounted the
LUN I’ve just created, referring to
http://docs.ceph.com/docs/mimic/rbd/iscsi-initiator-linux/.
I’m trying to do a HA testing:
1. Perform a WRITE test with DD command
2. Stop one ‘Activate’ iscsi-target node(ini 0), DD IO hangs over 25
seconds until iscsi-target switch to another node
3. DD IO goes back normal
My question is, why it takes so long for the iscsi-target switching? Is
there any settings I’ve misconfigured?
Usually it only take a few seconds to switch on the enterprise storage
products.
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
If you mean you shutdown the entire host, if so from your description
this is also running osds, so you also took out some osds serving io.
if a primary osd is not responding, clients io (in this case your iscsi
target) will block until ceph marks the osd down and issue a new epoch
map mapping the pg to another osd. This process is controlled by
osd_heartbeat_interval(5) and osd_heartbeat_grace(20) total 25 sec which
is what you observe. I do not recommend you lower them, else your
cluster will be over sensitive and osds could flap under load.
Maged
_______________________________________________
ceph-users mailing list
ceph-users@xxxxxxxxxxxxxx
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
