Performance

This menu entry includes the following options:

• Enable process watchdog (Enable watchdog service process, that ensures that process is always running)
• Process watchdog sets only internal errors (I GRID enviroments you can enable this setting and monitor for the internal error in logs and on email) introduced in ICP 4.3.0
• File cache limit file count (use this setting to set the cache limit by number of files)
• Hhash cache limit file count
• PostgreSQL VFS cache limit in MB
• Packet generator file cache limit (Limit the amount of cached files created by package generator)
• Packet generator minimum cache time in seconds
• Packet generator response cache limit
• Packet generator response cache time in seconds
  
• Oldest packet generator task record in seconds
• Keep maximum packet generator task records
• Packet generator report max execution time in seconds (Receive notification when creation process of an executable program exceeds set time).
• Packet generator internal error max execution time in seconds (Warning that creation process has exceeded the set time)
• Cache customizations in memory
• Enable indirect customization IDs
• Automatic database rebuild (default value: yes, set whether server should just stop or go into automatic rebuild of tables)
• Limit GRID error status updates in seconds
• Limit GRID replication status updates in seconds
  
• Limit GRID task status updates in seconds
• Software policy filter cache max entries

Watchdog

Watchdog is internal process that check ISL Conference Proxy threads to detect any deadlock. Default settings are that watchdog will restart ISL Conference Proxy in case any thread pool does not respond in specified time. Timeouts are set:

• 300 seconds for threads using disk IO
• 60 seconds for threads using only CPU and network IO
• 1800 seconds and 3600 seconds for special tasks like NTP and GeoIP.

Watchdog thread pools being monitored

Deadlock timeout ... 300s
Core cpu
Core main
Core dns
Core smtp
Core http_events
Core xmlmsg
Core disk
Core db
Core error_report
Core public_web
Core db_callback
Core file_check
Core send_in_memory
Core cpulow
Core rdns
Core alog
Core trans
Core wakeonlan
Core netgrid
Core netglock
Core dtgridsync
Core dt
Core swalrecv
Core NetMt TCP connect
Core NetMt resolve
DNS refresh zones
ISL AlwaysOn main
ISL AlwaysOn rpc
ISL AlwaysOn file
ISL AlwaysOn XMLMSG
ISL AlwaysOn DB flush
ISL AlwaysOn actions
ISL AlwaysOn refresh
ISL Groop messages
ISL Groop XNetMTInterface
ISL Light Netglock
ISL Light AUX
ISL Light DB events
ISL Pronto authenticator
ISL Pronto query

Deadlock timeout ... 60s
Core netcpu
Core netssl
Core netgridssl
Core NetMt TCP/IP events <NUM>
Core NetMt TCP accept <NUM>
ISL Pronto main

Deadlock timeout ... 1800s
GeoIP load

Deadlock timeout ... 3600s
NTP sync
NTP resolver

Watchdog and single server

You should leave default setting as are. Watchdog will restart ISL Conference Proxy in case any thread pool does not respond in specified time.

Watchdog and GRID environment

In GRID environment you can disable normal watchdog operation which will restart ISL Conference Proxy when deadlock occurs by enabling setting "Process watchdog sets only internal errors". Once the setting will be set to yes watchdog will only set "Internal error occurred: Deadlock in: <thread pool>" which causes load balance penalty for the server with the error.

Watchdog will not write any log line or send emails to system administrators. When disabling ISL Conference Proxy restart by enabling "Process watchdog sets only internal errors" system administrator needs to check ISL Conference Proxy administration regularly for errors.

Versions