-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Rhugga wrote:
> >We are having problems with rsh on 7.2 machines. We are doing massive >rsh connections on a 250-node cluster executing PVM jobs. We reach a >point where we get the following error: (single machine test) > >poll: protocol failure in circuit setup > >It happens around 500 or so connections. I wrote test scripts that >spawned rsh connections in sequence doing a simple 'uname -S' and >found that near 500 connections this error would occur. Note: It is >not always the same amount, it may be 497, 495, 499, but never 500. > >Is there a limit on open sockets or is there a limitation in in.rshd? >These machines are otherwise idle when these tests are being ran. My instinct would be to point the finger at xinetd. inetd's performance in this regard was known to suck; I don't know how xinetd compares, but I'd suspect it. I won't presume to know what's best for your situation, but there are usually better options than rsh -- ssh being the obvious one, and it's a lot faster than you might think, if you use a good algorithm such as blowfish. Another option for rshd might be to use a more robust services daemon such as tcpserver; http://cr.yp.to/ucspi-tcp.html. Just some thoughts. - -d - -- David Talkington PGP key: http://www.prairienet.org/~dtalk/0xCA4C11AD.pgp - -- http://setiathome.ssl.berkeley.edu/pale_blue_dot.html -----BEGIN PGP SIGNATURE----- Version: PGP 6.5.8 Comment: Made with pgp4pine 1.75-6 iQA/AwUBPIpK1r9BpdPKTBGtEQJELwCgonLjlO4FEp8wMilOBvFW6FpCQBQAoMdb lG43maJxSMB8N7QbaAdQBpCT =08Fm -----END PGP SIGNATURE----- _______________________________________________ Redhat-list mailing list [EMAIL PROTECTED] https://listman.redhat.com/mailman/listinfo/redhat-list