On AIX HA clusters (and probably other clusters also) when a node fails, the other node (in a two node cluster) take over the failed node's hostname, IP, etc. If you have a cron job, for example, that needs to run on a particular node, regardless of its state, the easiest way that I have found is to do a 'netstat -i' and grep for the node name. For example, I have a script I want to run every hour on node1. I make a crontab entry on both (or all) nodes in the cluster to run this script every hour. Node2 will start to run the script, encounter the HA check and exit. If node1 fails, node2 will take-over node1's name and IP, therefore the HA check will suceed because it now has node1's name and IP also.
sub HA_Check {
        @NETSTAT = split(' ', `netstat -i | grep node1`, 9999);
        if ($NETSTAT[3] !~ /node1/) {
                die;
        }
}

[download]

Replies are listed 'Best First'.
Re: HA node check
by Beatnik (Parson) on Dec 12, 2001 at 14:40 UTC
    if (((split(' ',`netstat -i | grep node1`))[3]) !~ /node1/) { die }

    might be a somewhat shorter. This, ofcourse, is puzzling me since you're grepping for lines with node1 and then are matching/dieing on lines that don't contain it.
    Update: Changed that split according to blakem's node below.

    Greetz
    Beatnik
    ... Quidquid perl dictum sit, altum viditur.
[reply]
[d/l]
[select]
      Splitting on / / is different from splitting on ' '. I believe that change would break the original program.... From split:
      As a special case, specifying a PATTERN of space (' ') will split on white space just as `split' with no arguments does. Thus, `split(' ')' can be used to emulate awk's default behavior, whereas `split(/ /)' will give you as many null initial fields as there are leading spaces. A `split' on `/\s+/' is like a `split(' ')' except that any leading whitespace produces a null first field. A `split' with no arguments really does a `split(' ', $_)' internally.
      Here is an example: 
      #!/usr/bin/perl -wT
use strict;

$_ = "   a   b   c   ";

print "split():    $_\n" for split();
print "split(' '): $_\n" for split(' ');
print "split(/ /): $_\n" for split(/ /);

__END__

=head1 OUTPUT

split():    a
split():    b
split():    c

split(' '): a
split(' '): b
split(' '): c

split(/ /): 
split(/ /): 
split(/ /): 
split(/ /): a
split(/ /): 
split(/ /): 
split(/ /): b
split(/ /): 
split(/ /): 
split(/ /): c

      [download]

      -Blake

[reply]
[d/l]
[select]
      Thats the whole point :) The netstat -i on node1 will return something like 
      Name  Mtu   Network     Address            Ipkts Ierrs    Opkts Oerrs 
+ Coll
en0   1500  link#2      0.4.ac.3e.65.22   169651762     0 150335201   
+  0     0
en0   1500  172.22.6    node1             169651762     0 150335201   
+  0     0
en2   1500  link#3      0.4.ac.3e.15.54    1387003     0  1389752     
+0     0
en2   1500  172.22.7    node1hb           1387003     0  1389752     0
+     0
en1   9000  link#4      0.4.ac.7c.95.ec   232558546     0 76198992    
+ 0     0
en1   9000  172.22.5    node1r          232558546     0 76198992     0
+     0
<loop back interfaces snipped>

      [download]
      and a netstat -i on node2 will return something like 
      
Name  Mtu   Network     Address            Ipkts Ierrs    Opkts Oerrs 
+ Coll
en0   1500  link#2      0.4.ac.3e.65.4a   36127938     0 37211569     
+0     0
en0   1500  172.22.6    node2             36127938     0 37211569     
+0     0
en2   1500  link#3      0.4.ac.3e.65.c6    1412752     0  1364210     
+0     0
en2   1500  172.22.7    node2hb           1412752     0  1364210     0
+     0
en1   9000  link#4      0.4.ac.7c.97.2f   76199094     0 232558588    
+11     0
en1   9000  172.22.5    node2r          76199094     0 232558588    11
+     0
<loop back interfaces snipped>

      [download]
      Where node? is the primary interface, node?hb is the heartbeat link and node?r is the redundant link that node2 will use to pretend to be node1 in the event of node1 failing, in which case interface en1 will no longer have a name of node2r but node1. Sooo, if node1 is up the test will fail on node2 (which is what we want) but if node1 is dead the test will succeed which, again, is what we want. I hope that makes more sense now.
[reply]
[d/l]
[select]

Back to Cool Uses for Perl