Appserver Side
Plugin Side
AppServer Side
→ Max WebContainer threads, set on
webcontainer
→ ConnectionIOTimeOut, set on
HTTP Transport
Time to wait when trying READ
data during request.
Default : 5 Seconds
→ ConnectionKeepAliveTimeout, “
Time to wait for next request on
a KeepAlive Connectio.
Default : 5 Seconds
Note: Ignore if
MaxKeepAliveConnections is ZERO.
→ MaxConnection Backlog, “
Maximum Connections queued up
before plugin refuses more connection requests.
Default : 512
Example: 50+512=562 concurrent
requests from plugin on port 9080 .
Here 50 are in application server
and rest are waiting in BACKLOG's FIFO queue in OS Kernel.
If 563rd
request comes to plugin, it will get rejected and will get an
ETIMEOUT error in http_plugin.log
Upperlimit for the MaxConnectionBacklog setting in the HTTPTransport
properties is limited by the Global OS Settings
If OS level backlog is 256,
HTTPTransport backlog is 512, The Effective backlog becomes 256.
→ MaxKeepAliveConnections, “
To Enable reuse of Http
Connections, tht already established b/w plugin & HttpTransport.
It prevents the overhead of
creating new requests.
Analogous to JDBC Connection
Pool.
Default : 90% of WC ThreadPool
→ MaxKeepAliveRequests, “
Maximum number of requests which
can be processed on a single KeepAlive Connection.
Default : 100
Ignore if MaxKeepAliveConnections
is ZERO.
Plugin (WebServer) Side
- OS Level .
- WebSphere Specific in plugin-cfg.xml .
Parameters that play major role
in failover and peak load scenarios.
a) OS Parameter : Tcp/Ip Timeout
when the tcp/ip client is not able to communicate with a tcp/ip
server in the time specified by tcp/ip timeout, the request is
aborted.
b) plugin-cfg.xml parameter :
ConnectionTimeout
If No ConnectionTimeout specified, the plugin performs a blocking
content in which the plugin sits until an operating system timesout
and allows the plugin to mark server unavailable
c) plugin-cfg.xml parameter :
RetryInterval
No comments:
Post a Comment