Issue 1267 - Provide a way to run some command on multiple hosts

run.py

 #!/usr/bin/env python
 # coding: utf-8
 
 import sys
 import os
 import re
 import subprocess
 import getopt
 import yaml
 
@@ skipping 78 matching lines @@
   monitoringConfig = readMonitoringConfig()
   if not monitoringConfig:
     print >>sys.stderr, "Failed to parse monitoring configuration"
     return [[], []]
   # Extract hosts and groups from monitoring config
   return processNode(monitoringConfig)
 
 def resolveHostList(hosts):
   validHosts, validGroups = getValidHosts()
   if not validHosts:
-    print "Warning: No valid hosts found, not validating"
+    print >>sys.stderr, "Warning: No valid hosts found, not validating"
     return hosts
 
   result = set()
   for param in hosts:
     if param in validGroups:
       for host in validGroups[param]:
         if host == '*':
           result = result | validHosts
         else:
           result.add(host)
@@ skipping 14 matching lines @@
   else:
     subprocess.check_call(command)
 
 if __name__ == "__main__":
   user, hosts, ignore_errors, args = parseOptions(sys.argv[1:])
   selectedHosts = resolveHostList(hosts)
   if len(selectedHosts) == 0:
     print >>sys.stderr, 'No valid hosts or groups specified, nothing to do'
     sys.exit(0)
   for host in selectedHosts:
-    print 'Running on %s...' % host

[mathias, 2014/08/25 16:21:25]

If all logging would be printed to >>stderr (just like it's intended to) instead
of polluting >>stdout, one could re-use the output of the command properly.

If the host information is required, one could prefix each line instead of
printing host notifications (triggered by a -p flag, for example), which is much
easier to process by scripts (|piped or via log file) or even manually
(shell/editor) later on.

This behavior could even be kept when e.g. deciding to implement parallelization
(similar to Fabric, see http://docs.fabfile.org/).

[Wladimir Palant, 2014/08/29 22:11:15]

I'm not usually using scripts on the output. And there might not even be any
output - I simply want to see the current progress when running a command on
twenty servers.
It's not that easy IMHO. Piping command output in Python while the command is
running requires a second thread - I'm not ready to go there for this simple
tool. An external tool like awk could do it as well, this adds a dependency
however. In the end, maybe I can simply pipe it into a second Python process -
but that's something to consider when I actually need this functionality.
I don't think I want to go there just yet :)

+    print >>sys.stderr, 'Running on %s...' % host
     runCommand(user, host, args, ignore_errors=ignore_errors)