Re: BDB Index corruption (ITS#2323)

[leifj@it.su.se]

Howard Chu wrote:

>Your trace does not provide any information about the problem you're reporting.
>Since your search used objectclass=posixGroup as its filter, the server never
>touched the uid index on this search.
>  
>
True, I sent you the wrong bdb-file... The effect is real however.

>The results from the search in your trace appear correct. The
>bdb_search_candidates function returned a list of 25 candidates for the
>objectclass=posixGroup search, out of the 183 immediate children of your search
>base. All 25 of those candidates matched the filter and were returned to the
>client. There is no indication that the server is processing any incorrect or
>unnecessary entries here.
>  
>
That is not the issue. The issue is that the search which normally takes 
< 1s takes 20-30 s...
I thought it happened because the index got corrupted and the DIT got 
traversed but I may
have been misstaken in my analysis.

>In the log, the lines:
>
>bdb_idl_fetch_key: [b40d1940]
><= bdb_index_read: failed (-30991)
>
>are normal. This is a search for entries of objectclass "referral" that back-bdb
>does for internal processing. The fact that it fails just means there are no
>referral objects in the database.
>
>  
>
ok, my misstake. Maybe we should look at this next week in SFO? I can 
show you hands
on what happens.

       Cheers Leif