11-21-2016 01:48 PM
On Sunday, I did a simple shut down and restart of the system. When it came up I saw a number of SAS processes start (couldn't tell what) and I started getting alerts.
Webapp CPU Time in Garbage Collection >30% on the mid-tier (tc Runtime SASServer1_1 ) I got this alert every 25 minutes for about 2 hours.
The mid-tier also was pegged at 99.6% CPU usage for about 30 minutes.
on the metadata server I got a bunch of TCP attempt fails right as the SAS system started (45.3 in one allert)
on the compute server was very much the same (46.3 in one allert)
This seem odd. Any suggestions. btw (9.4 M3 windows 2008 r2)
11-22-2016 03:43 AM
When a SAS system (in your case compute, metadata and midtier) starts, the web application server initiates all configured webapps. This will create heavy load and, depending on the power of the midtier server, take considerable time. Half an hour startup time for the tomcat(s) (or jboss) is not out of the norm.
I'd consider a startup sequence to make sure that unnecessary waits for not-yet-ready services are avoided:
- metadata server
- compute sever
Wait for the start of the metadata server process to finish before starting the compute server. Wait for the compute services (Spawner, OLAP) to finish startup before starting the Midtier.
Since I have it all on one machine, I have ordered it that way in /etc/inittab and inserted wait stages to allow time-consuming processes to finish their startup in the background before the next component starts.