Description of problem ====================== When machines are configured during *Create Cluster* task, the process is not done in parallel. For example: when packages are installed on one machine, the other machines are waiting and nothing is happening on them. Without fixing this BZ, it would not be possible to use USM to create a production sized clusters. Version-Release number of selected component ============================================ osd machine: ceph-0.94.5-9.el7cp.x86_64 ceph-common-0.94.5-9.el7cp.x86_64 ceph-osd-0.94.5-9.el7cp.x86_64 rhscon-agent-0.0.3-3.el7.noarch mon machine: ceph-0.94.5-9.el7cp.x86_64 ceph-common-0.94.5-9.el7cp.x86_64 ceph-mon-0.94.5-9.el7cp.x86_64 rhscon-agent-0.0.3-3.el7.noarch usm server machine: ceph-0.94.5-9.el7cp.x86_64 ceph-ansible-1.0.1-1.20160307gitb354445.el7.noarch ceph-common-0.94.5-9.el7cp.x86_64 redhat-ceph-installer-0.2.3-1.20160304gitb3e3c68.el7.noarch rhscon-ceph-0.0.6-14.el7.x86_64 rhscon-core-0.0.8-14.el7.x86_64 rhscon-ui-0.0.23-1.el7.noarch How reproducible ================ 100 % Steps to Reproduce ================== 1. Prepare machines (following usm documentation) for USM. Allocate at least 3 machines for monitor machines (no extra disks), and 4 machines (each with at least 2 extre disk for ceph OSDs) 2. In USM web interface, accept all machines. 3. Use *Create Cluster* wizard to setup a cluster, use all machines prepared in the step #1. 4. Check what is happening on each machine (ssh there and monitor cpu, disk and memory usage, run `top` there). 5. Wait for the process to finish. Actual results ============== The setup takes too long because setup is running only on the single machine at a time. This may also lead to the timeout caused failure of the Create Cluster task, but this is not a concern of this BZ. Expected results ================ There is some window of machines which are processed at the same time, and the admin can configure the size of this window. We should also evaluate if full parallel installation on all nodes is feasible. Additional info =============== The actuall component concerned with this may be different (devs told me that this may be a ceph-installer problem as well) - dev should proper reevaluate component of this BZ during investigation.
Additional information ====================== From the task details page, we can see that it took about 4 minutes to install packages on a machine, but since all were processed one by one, this process takes too long and would not scale for actual sized clusters with hunderds of machines: ~~~ Installing packages Mar 18 2016, 08:34:14 PM Installed packages on dhcp-126-80.lab.eng.brq.redhat.com: Mar 18 2016, 08:39:06 PM Installed packages on dhcp-126-85.lab.eng.brq.redhat.com: Mar 18 2016, 08:43:10 PM Installed packages on dhcp-126-81.lab.eng.brq.redhat.com: Mar 18 2016, 08:46:08 PM Installed packages on dhcp-126-84.lab.eng.brq.redhat.com: Mar 18 2016, 08:49:21 PM Installed packages on dhcp-126-82.lab.eng.brq.redhat.com: Mar 18 2016, 08:50:09 PM Installed packages on dhcp-126-79.lab.eng.brq.redhat.com: Mar 18 2016, 08:53:41 PM ~~~
This product is EOL now