dmcache cannot handle situation where lvmcache is not configured

[happz]

Without having any LVM cache set up command pcp dmcache leads to a traceback:

[root@foo ~]# pcp dmcache
Traceback (most recent call last):
  File "/usr/libexec/pcp/bin/pcp-dmcache", line 166, in <module>
    manager.run()
  File "/usr/lib64/python2.7/site-packages/pcp/pmcc.py", line 639, in run
    self._printer.report(self)
  File "/usr/libexec/pcp/bin/pcp-dmcache", line 137, in report
    max_lv = max_lv_length(group)
  File "/usr/libexec/pcp/bin/pcp-dmcache", line 82, in max_lv_length
    return len(max(lv_names, key=len))
ValueError: max() arg is an empty sequence
[root@foo ~]# 

Running pminfo -f dmcache reports (correctly) no values for dmcache metrics. The issue lies in formatting code, method max_lv_length expects there are always some data for dmcache available, and does not handle situation there are no data for key dmcache.cache.used. Very direct patch leads to a correct output (header + repeating lines of "No values available") but maybe a more sophisticated patch would be more appropriate.

diff --git a/src/pcp/dmcache/pcp-dmcache.py b/src/pcp/dmcache/pcp-dmcache.py
index a606eb1..ac243ca 100755
--- a/src/pcp/dmcache/pcp-dmcache.py
+++ b/src/pcp/dmcache/pcp-dmcache.py
@@ -78,6 +78,8 @@ def cache_dict(group, metric):
 def max_lv_length(group):
     """ look at the observation group and return the max length of all the lvnames """
     cache_used = cache_dict(group, 'dmcache.cache.used')
+    if not cache_used:
+        return 0
     lv_names = cache_used.keys()
     return len(max(lv_names, key=len))

[goodwinos]

thanks, pulled this in for 3.11.5. There is still an issue if the 'dm' PMDA is not ./Install'd, but when it is this patch improves the error handling.