diff options
Diffstat (limited to 'roles/ands_monitor/templates/scripts/check_server_status.sh.j2')
-rwxr-xr-x | roles/ands_monitor/templates/scripts/check_server_status.sh.j2 | 11 |
1 files changed, 11 insertions, 0 deletions
diff --git a/roles/ands_monitor/templates/scripts/check_server_status.sh.j2 b/roles/ands_monitor/templates/scripts/check_server_status.sh.j2 index caa63ce..b02f031 100755 --- a/roles/ands_monitor/templates/scripts/check_server_status.sh.j2 +++ b/roles/ands_monitor/templates/scripts/check_server_status.sh.j2 @@ -32,3 +32,14 @@ if [ -z "$disks" -o "$disks" -ne 0 ]; then echo "Not all disks are online:" /opt/MegaRAID/storcli/storcli64 /c0 show | grep -P "(HDD|SSD)" | grep -v "Onln" fi + +ifaces=$(ovs-vsctl show | grep -oP "could not open network device\s*\Kveth[a-f0-9]+" | wc -l) +if [ "$ifaces" -gt 50 ]; then + echo "Too many rogue interfaces ($ifaces) is registered on OpenVSwitch bridge. It could introduce large delays in pod scheduling..." +fi + +#Check various known problems +vssize=$(du -sm /var/log/openvswitch/ovs-vswitchd.log | cut -f 1) +if [ "$vssize" -gt 128 ]; then + echo "Current OpenVSwitch log is over $vssize MB. It could indicate some severe problems in pod networking..." +fi |