Closed ssinyagin closed 6 years ago
also garbage here, I just took the latest as-set object from ripe.db:
whois -h whois.mydomain:5043 AS21011:AS-CLIENTS
I re-downloaded fresh databases from radb.net and ripe.net, but the issue is still the same (the garbage output is different now)
what does the original object look like
Also garbage output:
$ whois -h whois.mydomain:5043 220.108.77.in-addr.arpa
The original is a normal as-set object, you can query them from RIPE or RADB
Do you also have the problem if you add to your configuration
irr_database ripe filter routing-registry-objects|route6
Do you need inetnum information at all?
Will give it a try. I need only AS sets and numbers, and routes.
I think most IRRd users run with that filter I pasted because they only need a subset of data, like you and me. It may be that the bug is not exposed with that filter enabled
with this filter, the memory footprint is much less than before, but the search for AS-DYNANET still produces garbage:
irr_database ripe mirror_host nrtm.db.ripe.net 4444
irr_database ripe mirror_protocol 3
irr_database ripe mirror-access 27
irr_database ripe max_journal_bytes 134217728
irr_database ripe filter routing-registry-objects|route6
irr_database radb mirror whois.radb.net 43
irr_database altdb mirror whois.radb.net 43
irr_database aoltw mirror whois.radb.net 43
irr_database arin mirror whois.radb.net 43
irr_database bboi mirror whois.radb.net 43
irr_database bell mirror whois.radb.net 43
irr_database canarie mirror whois.radb.net 43
irr_database easynet mirror whois.radb.net 43
irr_database host mirror whois.radb.net 43
irr_database jpirr mirror whois.radb.net 43
irr_database level3 mirror whois.radb.net 43
irr_database nestegg mirror whois.radb.net 43
irr_database nttcom mirror whois.radb.net 43
irr_database openface mirror whois.radb.net 43
irr_database ottix mirror whois.radb.net 43
irr_database panix mirror whois.radb.net 43
irr_database radb mirror whois.radb.net 43
irr_database reach mirror whois.radb.net 43
irr_database rgnet mirror whois.radb.net 43
irr_database risq mirror whois.radb.net 43
irr_database rogers mirror whois.radb.net 43
irr_database tc mirror whois.radb.net 43
$ whois -h whois.mydomain:5043 -- '!mas-set,AS-DYNANET'
A736
ed: 2006-02-06T09:14:35Z
source: RIPE
remarks: ****************************
remarks: * THIS OBJECT IS MODIFIED
remarks: * Please note that all data that is generally regarded as personal
remarks: * data has been removed from this object.
remarks: * To view the original object, please query the RIPE Database at:
remarks: * http://www.ripe.net/whois
remarks: ****************************
% Tags relating to '81.166.124.0 - 81.166.124.255'
% RIPE-USER-RESOURCE
inetnum: 194.122.142.248 - 194.122.142.255
netname: OTS60896
descr: Prevent Vochtwering bv
descr: VIANEN UT
country: NL
admin-c: DUMY-RIPE
tech-c: DUMY-RIPE
statuC
Can you check which version or commit of IRRd introduces this problem? I cannot reproduce on IRRd 2.3.10
ok, I'll try. The problem is reproduced with v3.0.8. Downgrading further now.
I reproduced it with v2.3.10, so it must be something in fresh dump from ftp.ripe.net
maybe there is some strange character somewhere
in
grep as-set: local/var/spool/irr_database/ripe.db
the failure is somewhere between lines 3000 and 8000. I'll make a script that finds the place where it starts failing.
Here, after a non-ASCII symbol, it went crazy:
inet6num: 2a01:4f8:c17:3231::/64
netname: STREAMABC-UG-HAFTUNGSBESCHRAENKT
descr: streamABC UG (haftungsbeschränkt)
country: DE
admin-c: TM6822-RIPE
tech-c: TM6822-RIPE
status: ASSIGNED
notify: ripe-mntner@hetzner.de
mnt-by: HOS-GUN
created: 2016-03-03T02:05:35Z
last-modified: 2016-03-03T02:05:35Z
source: RIPE
I installed a fresh dump from ripe.net, and disabled mirroring, and then all objects before this one are fine, and the objects after that one are generating garbage.
Please adapt the *.db reading procedure to accept non-ASCII symbols. This one is a Latin1 symbol, and I guess there might also be Unicode symbols eventually.
I described the problem to RIPE DBM, and they say it's up to IRRd developers to fix this issue.
I'd welcome a patch :)
I'm afraid it would take too much time for me to explore the depths of the code
yes, this is a challenge :(
It turned out that this bug is not Latin1 related in any way but is all about DB size: RIPEdb has grown over 4Gb.
See #48 for a patch.
woah.. great find! :)
I'm running a mirror of RIPE database on IRRd from git master. Queries on some RIPE objects produce garbage output (I found so far only one such object).
Configuration:
Query and result: