Packstack install fails on machine with many cores; --service-workers fixes

[W. Michael Petullo]

I am using Packstack 21.0.0 to install OpenStack Zed on CentOS Steam 9. I ran the following command:

packstack --allinone --os-neutron-l2-agent=openvswitch --os-neutron-ml2-mechanism-drivers=openvswitch --os-neutron-ml2-tenant-network-types=vxlan --os-neutron-ml2-type-drivers=vxlan,flat --provision-demo=n --os-neutron-ovs-bridge-mappings=extnet:br-ex --os-neutron-ovs-bridge-interfaces=br-ex:eno12399np0

The computer has 128 cores and provides 256 threads. The computer also has 512 GB of memory.

The install took a very long time, and it eventually fails due to a gateway timeout. I found that the HTTP interface/horizon is nearly unusable due to delays, and it furthermore commonly fails entirely due to gateway timeouts. Of course, the install failed, so the system is not fully functional in addition to the delays and timeouts.

I have found that passing "--service-workers=16" to packstack, along with the options above, fixes this problem. With the service workers limited to 16 the install completes and the HTTP interface/horizon becomes responsive.

Should there be a default upper limit to the number of service workers?