YunoHost/package_check
1
0
Fork 0
mirror of https://github.com/YunoHost/package_check.git synced 2024-09-03 20:06:20 +02:00
Code Issues Projects Releases Packages Wiki Activity

Merge pull request #170 from YunoHost/revert-165-reduce-fail-timeout

Browse source 
Revert "Fail startup/network timeout earlier. Exit whole test suite."
This commit is contained in:
Alexandre Aubin 2024-08-30 17:56:15 +02:00 • committed by GitHub
commit daa1e138cf
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 61 additions and 38 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

85
lib/lxc.sh
View file

@ -57,12 +57,7 @@ LXC_CREATE () {
log_critical "Failed to create the new LXC :/" log_critical "Failed to create the new LXC :/"
fi fi
# The first time around when we create the VM, if it fails, we want to abort early _LXC_START_AND_WAIT $LXC_NAME
# instead of keeping the timeout game until the end of times
if ! _LXC_START_AND_WAIT $LXC_NAME; then
log_critical "Fatal error starting container. See logs above."
fi
sleep 3 sleep 3
if ! $lxc exec $LXC_NAME -- test -e /etc/yunohost if ! $lxc exec $LXC_NAME -- test -e /etc/yunohost
@ -225,42 +220,70 @@ LXC_RESET () {
_LXC_START_AND_WAIT() { _LXC_START_AND_WAIT() {
# Try to start the container
restart_container()
{
LXC_STOP $1
$lxc start "$1"
}
# Try to start the container 3 times.
local max_try=3
local i=0
while [ $i -lt $max_try ]
do
i=$(( i +1 ))
local failstart=0
# Wait for container to start, we are using systemd to check this, # Wait for container to start, we are using systemd to check this,
# for the sake of brevity. # for the sake of brevity.
for j in $(seq 1 5); do for j in $(seq 1 10); do
log_debug "Start container attempt $j" if $lxc exec "$1" -- timeout 30 systemctl isolate multi-user.target >/dev/null 2>/dev/null; then
if timeout -k 10 4 $lxc exec "$1" -- timeout 4 systemctl isolate multi-user.target >/dev/null 2>/dev/null; then
break break
fi fi
if [ "$j" == "5" ]; then if [ "$j" == "10" ]; then
log_error 'Failed to start the container.' log_debug 'Failed to start the container ... restarting ...'
$lxc info --show-log $1 failstart=1
return 1
fi
done
log_info "Container started successfully. Now waiting for internet access." restart_container "$1"
# Wait for container to access the internet
for j in $(seq 1 5); do
log_debug "Connect container internet attempt $j"
# Note: Sometimes this uses X00% CPU and never times out, so we use timeout SIGKILL
if timeout -k 10 4 $lxc exec "$1" -- timeout 4 curl -s http://wikipedia.org > /dev/null 2>/dev/null; then
break
fi
if [ "$j" == "5" ]; then
log_error 'Failed to access the internet'
$lxc info --show-log $1
return 1
fi fi
sleep 1s sleep 1s
done done
log_debug "Container fully started"
# Wait for container to access the internet
for j in $(seq 1 10); do
if $lxc exec "$1" -- timeout 10 curl -s http://wikipedia.org > /dev/null 2>/dev/null; then
break
fi
if [ "$j" == "10" ]; then
log_debug 'Failed to access the internet ... restarting'
failstart=1
restart_container "$1"
fi
sleep 1s
done
# Has started and has access to the internet
if [ $failstart -eq 0 ]
then
break
fi
# Fail if the container failed to start
if [ $i -eq $max_try ] && [ $failstart -eq 1 ]
then
log_error "The container miserably failed to start or to connect to the internet"
$lxc info --show-log $1
return 1
fi
done
sleep 3
LXC_IP=$($lxc exec $1 -- hostname -I | cut -d' ' -f1 | grep -E -o "\<[0-9.]{8,}\>") LXC_IP=$($lxc exec $1 -- hostname -I | cut -d' ' -f1 | grep -E -o "\<[0-9.]{8,}\>")
} }