Hello all,
After a couple weeks of shutting down my workstation over the weekend, I noticed that the server didn't remove it from the "blackout" list. After looking into it (I won't tell you how much time it took...*grrrr*), I found that the ping command wasn't returning a positive error status, when using the "-s" parameter. Thus:
/usr/sbin/ping -s <non-responding IP> 56 1
...
echo $?
0 <-- should have been positive
I verified this on Solaris 2.6 and 2.7. The ping command works correctly on Solaris 2.8 (and probably 2.9). Couldn't find anything on sunsolve or elsewhere. The result was that the ping test succeeded. Then the xfer routines kicked in which failed with "Got fatal error during xfer (Connection to <host> failed)" messages. The quick fix was to just use "ping $host 1" as the test and while it doesn't allow for RTT, the code in Lib.pm (CheckHostAlive) will return zero if it can't match any "time" string. Craig, ever think about using the Perl "ping" library? Just wondering, mind you.
I also noticed an oddity, perhaps by design (Craig?). Even if you haven't defined a host in the "conf/hosts" file, you can still use BackupPC_dump to do a backup. Several conditions must be met, mind you (share, share password, etc.). The host directory must exist, also. This could be interesting for quick tests but the code seems to only use the "server" config.pl even if you have created one in the host's directory. I assume this is due to the fact that the daemon doesn't know about the host and any attempt to access {Conf} will always point to the server config. Perhaps a check for config validity before allowing a backup?
Just fyi....
- Wade
