Table of Contents

Name

crawl - self-ping utility (for crawling myrinet routes on Cplant)

Synopsis

crawl  node-id

OR

crawl  route-byte-sequence

Description

crawl must be run as the superuser.

crawl is a self-ping utility for myrinet on Cplant. It is run on a service or compute node to test each hop in a myrinet route from the local node to a specified desination node. It takes as argument the node id of the destination machine or a space-separated list of route bytes, each of the form 0xST with S and T hex digits.

crawl is a perl script that translates its local node id to a machine name by calling pnid2hname. In the case of a specified destination node it uses this name along with the specified destination node id to parse a route sequence out of the directory of Cplant route files. Then it invokes the troute(1) utility. troute in turn makes ioctl() calls to the portals and rtscts devices.

crawl performs a self-ping test for each byte of the route on the path to the destination machine; each time a ping packet enters a new switch it immediately goes back out that port and retraces the route back to the node of origin.

crawl must be run as the superuser.

See Also


Table of Contents