How HTTP Requests are Processed in the Dispatcher Node

You may be aware by now that when you are calling 5<Instance Number>00 port, you are sending the request on dispatcher node's http port. The call is routed from that point to the server node where the application logic processes the request. This post briefly discusses what happens inside.

Let's assume that there is a flood of requests made to the port. The requests are queued up in the socket queue. These requests are listened to at the ServerSocketListener for http port. The socket queue and the listeners are handled by the Ports Manager. The HTTP Provider service provides the port and other information such as keepalive timeout and the number of listeners (AcceptingThreadsCount)

The request picked up by the listener will be in the waitingTasks queue (big pile of files in the diagram) until it is picked up by a System Thread for processing. The Connection Manipulator wraps the request with a TCPRunnableConnection and sends it to the HTTP protocol processor.

The Cluster Manager then sends the request (through Session Communication) to the Server node which ultimately processes the request.


The Cluster Manager is also responsible for managing cluster communication via Message Server or without involving MS for lazy communication. These aspects are not discussed here.

In order to scale the abilities to handle HTTP requests up, the following parameter play a vital role: AcceptingThreadsCount, SocketQueue, KeepAliveTimeout and System Threads. An increment in these parameter will obviously result in an increase in memory usage.

Comments

Post a Comment

Popular posts from this blog

OS/DB Migration - CMD. STR, TOC, EXT, R3load, DDLDBS.TPL and more

Image
With increasing need to improve performance or reduce costs or vanishing support for operating systems/databases it becomes inevitable to migrate to a better OS or DB. We feel take a quick look at the primary ingredients of the tools for OS/DB migration handled by R3SETUP. Tools - R3load, R3ldctl and R3szchk R3ldctl creates STRucture files, TemPLate files by reading the technical settings for table/indexes in SAP data dictionary. R3szchk calculates the database size object sizes (EXTent files) for the target system based on the STR files created by R3ldctl. The EXT and STR files can be later split into smaller groups. The most crucial tool R3load is responsible for unloading and loading data from the database. R3load uses the files created by R3ldctl and R3szchk and dumps the data into compressed binary format as .001, .002 .. files. It uses cmd files to achieve the task. If the cmd files are not present, it creates on its own. Files Command files*.cmd  are created i
Read more

Fixing Inconsistent Table - Table activation fails due to inconsistency between DD and DB

If you find a table to be available on the database, but not active in ABAP dictionary, try calling SE14 as DDIC user, enter the table name and choose Table > Reconstruct, Execute. If this method fails follow these steps: Call transaction code SE37 and execute the function module  DD_TABL_ACTM with single test On the next screen, do not change any other field except the following and set them as shown below: MODE: 12 NTAB_PUTSTATE: N TABNAME: <enter the correct table name> Click on Execute The export parameter ACT_RESULT should show 0. To verify if the inconsistency is fixed, call transaction SE14 > Extras > Database Object > Check and  SE14 > Extras > Runtime object > Check.
Read more

301 Redirect Using SAP Web Dispatcher

301 redirects are typically used to redirect accesses that include www to non-www or vice-versa. Example, when you call http://blogger.com, the call is redirected to www.blogger.com. An end-user may prefer to call a URL in any manner. You must ensure that all these calls are redirected to relevant URLs with content. SAP web dispatcher can be configured to send a redirect (301) to the client. To enable redirection add icm/HTTP/redirect_<xx> to the web dispatcher's profile file. The parameter uses the following syntax: icm/HTTP/redirect_<xx> = PREFIX=<URL prefix>[, FROM=<pattern for URL>, FROMPROT=<incoming protocol>, FOR=<pattern for host name:port>,TO=<new URL prefix>, PROT=<protocol>, HOST=<host>, PORT=<port number/name>] Let's say you want to redirect all the HTTP requests on mysource.com on 80 port to mytarget.com on HTTPS 443 port icm/HTTP/redirect_0 = PREFIX=/, FROM=*, FROMPROT=http, FOR=mysource.com:80,
Read more