Closed terrywbrady closed 4 weeks ago
Suggest testing on Stage by taking one inv host out of the load balancer, requesting to alter the other inv host to a larger instance type, and testing with the largest manifest we have to see where incremental increases in memory no longer help. Then we can determine what an optimal jvm memory configuration is.
Increased Inventory JVM for prod to 1.5GB Will take effect after Tuesday 6/4 patching
Some things to note with these exceptions:
Comments:
***** Nagios *****
Notification Type: PROBLEM
Service: uc3-mrt-inventory-prd_state_7x16 Host: uc3-mrtinv-prd01 Address: uc3-mrtinv-prd01.cdlib.org State: CRITICAL
Date/Time: Tue Jun 4 10:40:21 PDT 2024
Additional Info: HTTP CRITICAL: Status line output matched HTTP/1.1 200 - 986 bytes in 5.684 second response time
Nagios
Notification Type: PROBLEM
Service: uc3-mrt-inventory-prd_status-running_7x16 Host: uc3-mrtinv-prd01 Address: uc3-mrtinv-prd01.cdlib.org State: CRITICAL
Date/Time: Tue Jun 4 10:41:07 PDT 2024
Additional Info: HTTP CRITICAL: HTTP/1.1 200 - 986 bytes in 9.534 second response time
Nagios
Notification Type: RECOVERY
Service: uc3-mrt-inventory-prd_state_7x16 Host: uc3-mrtinv-prd01 Address: uc3-mrtinv-prd01.cdlib.org State: OK
Date/Time: Tue Jun 4 11:00:21 PDT 2024
less localhost_access_log.2024-06-04.txt
172.30.32.237 - - [04/Jun/2024:10:40:21 -0700] "GET /state?t=xml HTTP/1.1" 200 747 172.31.14.167 - - [04/Jun/2024:10:40:21 -0700] "GET /state HTTP/1.1" 200 855
172.31.14.167 - - [04/Jun/2024:10:41:07 -0700] "GET /state HTTP/1.1" 200 855
172.31.14.167 - - [04/Jun/2024:11:00:21 -0700] "GET /state HTTP/1.1" 200 855 172.30.32.237 - - [04/Jun/2024:11:00:21 -0700] "GET /state?t=xml HTTP/1.1" 200 747
Ashley's analysis suggests that increasing the JVM size will affect the swap issue spikes we see. We will increase the JVM on prod. Inventory from 1GB to 1.5GB on Monday and the reboot after patching on Tuesday will allow it to take effect.