[squid-users] Warning about "Invalid entries" in cache.log (rock-store)

[Alex Rousskov]

On 11/25/2015 11:22 PM, Tom Tom wrote:


> I'm running Squid 3.5.11 (Linux, 64Bit) with 16 workers and 4
> cache_dir's (rock) configured.
> 
> The 4 rock-caches where newly builded a few days ago. In the meantime,
> during squid-startup, I receive warnings like this:
> ...
> ...
> 2015/11/26 00:07:41 kid17| WARNING: Ignoring malformed cache entry.
> 2015/11/26 00:07:41 kid17| WARNING: Ignoring malformed cache entry.
> 2015/11/26 00:07:41 kid17| WARNING: Ignoring malformed cache entry.
> 2015/11/26 00:07:41 kid17| WARNING: Ignoring malformed cache entry.
> 2015/11/26 00:07:41 kid17| WARNING: Ignoring malformed cache entry.
> 2015/11/26 00:07:42 kid17| Finished rebuilding storage from disk.
> 2015/11/26 00:07:42 kid17|   2879999 Entries scanned
> 2015/11/26 00:07:42 kid17|      1092 Invalid entries.
> 2015/11/26 00:07:42 kid17|         4 With invalid flags.
> 2015/11/26 00:07:42 kid17|    158845 Objects loaded.
> 2015/11/26 00:07:42 kid17|         0 Objects expired.
> ...
> ...
> 
> 
> How critical are the "Invalid entries"? What's the reason for these?

Most invalid entries are the results of "normal" conflicts when multiple
workers are trying to store the same entry and/or dealing with aborted
transactions. Squid crashes and restarts may add quite a few of these,
for example.

Unfortunately, it is not possible to say whether your "invalid entries"
are critical -- there is not enough information in the above stats. A
good rule of thumb may be to ignore a _small_ portion of invalid entries.

If yo want to dig deeper, enable "47,2" (at least) debugging using
debug_options and look for "freeBadEntry" lines. They should contain the
reason(s) why Squid considers your entries invalid. There may be other
"invalid entry" causes, but IFAICT, all others should be accompanied by
level-1 WARNINGS and/or other error counters.


HTH,

Alex.