SUMMARY: problem with rcp

From: Jake Mahon (jakem@apex.com)
Date: Wed Jul 28 1993 - 01:39:49 CDT


[ My original question:
[ Our nightly backup routine copies a 400 Mb databse file to a
[ development machine. Both machines are runing 4.1.2 mostly
[ generic kernels (max users up'd). About one night out of five,
[ the copy will fail and cause a spurious reboot of the source
[ machine.
[
[ The sh script is simply /usr/ucb/rcp /ata/db remotemachine:/ata/db
[
[ The copies always seem 95% complete when the local machine reboots.
[ I have also seen this behavior when the rcp is called from the
[ command line, rather than via cron. I haven't yet been able to
[ catch any console error messages.
[
[ Can anyone suggest a cause or a solution for this rcp failure?

ray@isor.vuw.ac.nz (Ray Brownrigg) comments on rcp performance:
: ... I am sure you will find that if you run that
: command from the *receiving* machine instead of the "sending" machine,
: you will notice a substantial improvement in speed. The reason is
: something to do with NFS using a separate write for every 512-byte block
: at the receiving end, but NFS reads are much more efficient....

Although my timimg issues for starting and stopping the database servers
became stickier, I'm now initiating the rcp from the destination machine.
If I still get a spurious reboot (too soon to tell), at least it will
be my development machine crashing and not the production machine.

ROtt@Logicon.COM (Randy J. Ott) discusses rcp expansion of null bytes:
> When cp or rcp is involved, it expands these NULL bytes so that they occupy
> actual space on the file system. The 10001 byte file you cp or rcp will
> occupy 10001 bytes on the file system. What does this have to do with your
> problem? I'm not sure, but it might be possible that you are running out of
> disk space on the remote machine and it's rebooting for some reason or
> another. (I'm reaching on this one...)

Randy further points out this can be overcome w/ a tar|rsh pipe. I've
got lots of disk space, so didn't pursue this welcome suggestion.

rwolf@dretor.dciem.dnd.ca (Robert J Wolf) advises:
| Upgrade to Sun0s 4.1.3, install all nfs, ufs, tcp loopback, tmpfs patches.

This is good advice, I'm sure. Given the amount of overhead involved,
I need to revisit the apparently inevitable Solaris 2.x upgrade.



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:08:04 CDT