[Netarchivesuite-devel] Analysing first stage

[Bjarne Andersen]

My best guess would be an error on one of the harvster-servers. When finishing a job NAS creates the metadata-file and creates statistics based on the crawl.log and sends a return message to NAS. This return-message must have failed for some reason I don't quite know.

Can you see if all the domains were harvested in the same job ? (you can select from the seeds-column of the job-table to find jobs with those domains). Is all jobs in the entire harvest reported in NAS - either DONE or FAILED or is any jobs missing a status ?

best
Bjarne
________________________________________
Fra: sara.aubry at bnf.fr [sara.aubry at bnf.fr]
Sendt: 12. maj 2010 15:47
Til: netarchivesuite-devel at lists.gforge.statsbiblioteket.dk; Bjarne Andersen
Emne: Analysing first stage

Hello everyone,

Our first stage finally completed on Monday 10 and lasted longer than
expected (we started on April 14).
We collected 1000 URL on 1 673 094 domains, using 4 063  031 seeds.

Our main figures:
- Number of jobs: 497 (among which 17 failed and were resubmitted)
- Number of crawled URL: 294 387 919
- Number of crawled hosts: 3 234 464
- Number of ARC files: 107 173
- Size of compressed ARC files: 10,44 Tb

75 424 domains reached the max object limit and will be part of the second
stage we will launch on May 17.

While QAing our first stage and crossing figures, we noticed that 3 496
domains (all starting with an s) were
missing from the historyinfo table and were not harvested at all.

Looking quickly at the missing domain list, we first thought that NAS
didn't like porn sites :-) (many domain names with sex, sexy...) but the
list is larger than that.
There are no errors in the GUIApplication log file, nor in the DB log
file.

This is no big deal, only 0.2% of the entire seed list, and we will just
run a patch crawl.
But have you ever had this problem?

Best,

Sara




Avant d'imprimer, pensez ? l'environnement.