PackStack Quick Start - hpaluch/hpaluch.github.io GitHub Wiki
PackStack Quick Start
PackStack (formerly known as RDO) is quick installation for OpenStack provided by RedHat
Setup
- VM under Proxmox VE 6, 6GB of RAM, 2x core, 12GB disk space
- Tested VM OS (where PackStack is installed):
CentOS Linux release 7
WARNING! I don't recommend using CentOS 8 for these two reasons:
- shorter EOL (Dec 2021) than CentOS 7 EOL (June 2024) - see
- only CentOS Stream will be available in future:
I strongly recommend this (very good indeed!) guide as starting point:
- https://www.tecmint.com/openstack-installation-guide-rhel-centos/ Or official installation guide form:
- https://www.rdoproject.org/install/packstack/
However I rather used this one which allows access from existing network:
I just additionally enabled Heat (provisioning), Pointed cinder to
dedicated volume group stack-vg
, disabled Swift (S3
like service), disabled Ceilometer (Performance monitoring). You can
see my latest setup script here:
There is manual demo user setup on above page - I automated part (!) of setup here:
- https://github.com/hpaluch/openstack-examples/blob/master/packstack/setup_demo.sh But be warned that I use unique names to catch errors in parameters definition.
Notes:
- out of the box the demo user/vm can access Internet, thanks to provided NAT.
- However if you need to access VM from your physical network you need to:
-
create floating IP address on pool
external_network
(say192.168.0.39
in my example) -
associate it with running instance
-
you can directly access your demo (
cirros-1
from mysetup_demo.sh
script) running instance using Floating IP address (for example SSH to192.168.0.39
in my example):# key demo_kp.pem was created in HOME directory # by `setup_demo.sh` script on your PackStack server ssh -i demo_kp.pem [email protected]
-
Issues
After bare metal install (8GB RAM, AMD X-2 dual-core) of Train
release I noticed very bad performance.
When user demou
issued this trivial command it took around 8 seconds:
time openstack server list
# or nice --timing parameter
# from https://ask.openstack.org/en/question/99737/2-second-delay-on-openstack-api/
time openstack server list --timing
I just tried this command to verify that keystone caching is not enabled:
$ sudo keystone-manage doctor
WARNING: `keystone.conf [cache] enabled` is not enabled.
Caching greatly improves the performance of keystone, and it is highly
recommended that you enable it.
I verified that memcached
is already installed and running:
$ ps ax | grep memcached
1338 ? Ssl 0:01 /usr/bin/memcached -p 11211 -u memcached -m 797 -c 8192 -l 0.0.0.0 -U 11211 -t 2 >> /var/log/memcached.log 2>&1
11571 pts/0 R+ 0:00 grep --color=auto memcached
And tested with telnet:
$ sudo yum install telnet
$ telnet localhost 11211
...
Connected to localhost.^M
Escape character is '^]'.^M
stats
STAT pid 1338
STAT uptime 3891
STAT time 1608026790
...
END
quit
So I followed this page https://docs.openstack.org/keystone/latest/admin/configuration.html#configure-the-memcached-back-end-example
to enable Memcached backend for Keystone - I changed file /etc/keystone/keystone.conf
:
diff -u /etc/keystone/keystone.conf{.orig,}
--- /etc/keystone/keystone.conf.orig 2020-12-15 10:21:06.069000000 +0100
+++ /etc/keystone/keystone.conf 2020-12-15 11:06:58.117000000 +0100
@@ -374,6 +374,10 @@
[cache]
+# from https://docs.openstack.org/keystone/latest/admin/configuration.html#configure-the-memcached-back-end-example
+enabled = true
+backend = dogpile.cache.memcached
+backend_argument = url:127.0.0.1:11211
#
# From oslo.cache
And then restarted keystone (it is actually Python module run in Apache process):
sudo systemctl restart httpd
When I tried again time openstack server list
I got a bit better response - around 5 seconds.
Hmm a bit better but still quite bad.
Also note that Idle system (no running instance and no active commands) causes load average around 1.0
Resources
Last RedHat guide for Manual OpenStack installation - very useful if want to know how OpenStack works: