(safplus_console amsdbxmlprint timeout, safplus_info forest timeout) |
Latest revision as of 16:10, 28 March 2013
[edit] Symptom Statement
- The safplus_console command "amsdbxmlprint console" can hang.
- The safplus_info command "forest" times out.
[edit] Diagnosis
These commands can timeout if the rmem_default/max buffer in the kernel gets exceeded which results in packet drops. This will happen with a large model since the amsdbxmlprint command dumps the entire cluster configuration as XML data. If the "console" modifier is removed, the command should work since it dumps into file on the controller. The default rmem_default/max is around 128k.
[edit] Solution
Increase the rmem defaults through a proc write:
echo 10000000 > /proc/sys/net/core/rmem_default
echo 10000000 > /proc/sys/net/core/rmem_max
Now try the amsdbxmlprint console command and see it work. Of course, the cluster has to be up with SAFplus running.
Make it persistent across reboots with a /etc/sysctl.conf entry:
net.core.rmem_default = 10000000
net.core.rmem_max = 10000000