[prev in list] [next in list] [prev in thread] [next in thread] 

List:       inn-workers
Subject:    Re: bug in expire? corrupt history?
From:       Russ Allbery <rra () stanford ! edu>
Date:       2000-03-31 10:46:06
[Download RAW message or body]

Answering old mail, so chances are that this was resolved some time back,
but I wanted to make sure and get these messages taken off my list of
unanswered ones.

Paul Theodoropoulos <paul@atgi.net> writes:
> At 11:48 AM 01/14/2000 , Paul Theodoropoulos wrote:

>> this is a followup to my messsage of jan 12, regarding inn throttling on

>> Jan 11 17:25:05 nnrp innd: SERVER cant dbzstore 
>> [5896A8AD039F14210F44D65600BD43D8]@0: Invalid argument

> further followup. Since I generally fly by the seat of my pants, i
> decided to run sed on history, and remove the occurrances of that hash
> from the history file. i then ran makedbz, and came up - once again -
> with a fine 156meg history.pag.

> Here's partial output of grep before having run sed (full output was 118 
> lines):

> news-nnrp /usr/local/news/db% cat badmess
> [5896A8AD039F14210F44D65600BD43D8]      947745808~-~947745808
> [5896A8AD039F14210F44D65600BD43D8]      947745817~-~947745817
> [5896A8AD039F14210F44D65600BD43D8]      947745992~-~947745992
> [5896A8AD039F14210F44D65600BD43D8]      947746117~-~947746117
> [5896A8AD039F14210F44D65600BD43D8]      947746119~-~947746119
> [5896A8AD039F14210F44D65600BD43D8]      947746299~-~947746299
> [5896A8AD039F14210F44D65600BD43D8]      947746371~-~947746371
> [5896A8AD039F14210F44D65600BD43D8]      947746451~-~947746451

> I'm not sure how to interpret this. I guess I also don't understand why
> a hash is being used over and over again, but that is likely an
> incomplete understanding of the structure of history.

This looks very much like the hash overflow problem that we've received
scattered reports of with the tagged hash history file.  That seems to be
the cause of the duplicate hashes, as best as I could follow previous
discussion of this.  I'm hoping that these sorts of problems will
eventually become much easier to debug with a new history API and backend,
probably in 2.4....

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>

[prev in list] [next in list] [prev in thread] [next in thread]