Files
devstack-gate/slave_scripts/baremetal-deploy.sh
James E. Blair 7874c0cec0 Add bare metal deployment scripts.
Add ping.py so we can test when a host is up.  ping -c 2 -w 90 will
almost do the right thing, but will exit on a icmp host unreachable.
We need to continue to wait.  ping.py just waits for an answer to a
ping.

Parallelize building the compute nodes (but not the head node).

Kill ping process in ping.py.

Copy PIP cache to host before running.

Add syslog artifacts.
Remove known_host weirdness, depend on .ssh/config instead.

Change-Id: I65a12074cf80feb3ae69d15c81ba32a81566b22a
2011-10-19 20:59:08 -04:00

44 lines
1.2 KiB
Bash
Executable File

#!/bin/bash
set -x
#HEAD_HOST=${HEAD_HOST:-10.14.247.43}
#COMPUTE_HOSTS=${COMPUTE_HOSTS:-10.14.247.44,10.14.247.45}
WORKSPACE=`pwd`
mkdir -p logs
rm -f logs/*
cd `dirname "$0"`
echo "Jenkins: resetting hosts..."
for host in $HEAD_HOST ${COMPUTE_HOSTS//,/ }; do
scp -o lvm-kexec-reset.sh root@$host:/var/tmp/
ssh -o root@$host /var/tmp/lvm-kexec-reset.sh
sudo rm -f /var/log/orchestra/rsyslog/$host/syslog
done
# Have rsyslog reopen log files we rm'd from under it
sudo restart rsyslog
# wait for the host to come up (2 ping responses or timeout after 5 minutes)
echo "Jenkins: Waiting for head host to return after reset..."
if ! timeout 300 ./ping.py $HEAD_HOST; then
echo "Jenkins: ERROR: Head node did not come back up after reset"
exit 1
fi
echo "Jenkins: Pre-populating PIP cache"
for host in $HEAD_HOST ${COMPUTE_HOSTS//,/ }; do
scp -r ~/cache/pip root@$host:/var/cache/pip
done
echo "Jenkins: Executing build_bm_multi.sh."
cd ~/devstack
source ./functions.sh
cache_images ~/devstack/files
bash build_bm_multi.sh
for host in $HEAD_HOST ${COMPUTE_HOSTS//,/ }; do
cp /var/log/orchestra/rsyslog/$host/syslog $WORKSPACE/logs/$host-syslog.txt
done