new data not imported?

[job]

It appears there is a difference between what is stored on disk, and what is served up through the daemon. Perhaps I configured it wrong?

Every night a new rpki.db file is created, and 1 is added to the value in RPKI.CURRENTSERIAL, these are copied into place.

irrd@irrd01:~/routing_sources/rpki$ cat /home/irrd/dbs/RPKI.CURRENTSERIAL
186
irrd@irrd01:~/routing_sources/rpki$ grep changed: /home/irrd/dbs/rpki.db | sort | uniq
changed:    job@ntt.net 20190127

hanna:~ job$ whois -h irrd01.ip.gin.ntt.net -- -s RPKI -M 185.0.0.0/8  | grep changed | sort | uniq
changed:        job@ntt.net 20190115

This is the configuration, and RPKI is listed under sources_default: too:

        RPKI:
            nrtm_access_list: generic_nrtm_access
            import_serial_source: file:///home/irrd/dbs/RPKI.CURRENTSERIAL
            import_source: file:///home/irrd/dbs/rpki.db

[job]

on disk serial is 186, on DB i see 181 as latest

                  pk                  | source | serial_oldest_seen | serial_newest_seen | serial_oldest_journal | serial_newest_journal | serial_last_export | force_reload | last_error | last_error_timestamp |            created            |            updated
--------------------------------------+--------+--------------------+--------------------+-----------------------+-----------------------+--------------------+--------------+------------+----------------------+-------------------------------+-------------------------------
 492b3f69-3f2b-4bd2-8f0a-c985d277b70c | RPKI   |                181 |                181 |                       |                       |                    | f            |            |                      | 2019-01-16 20:43:12.670088+00 | 2019-01-16 20:50:16.954105+00

[job]

I set force_reload to true, and this triggered an actual reload:

hanna:~ job$ whois -h irrd01.ip.gin.ntt.net -- -s RPKI -M 185.0.0.0/8  | grep changed | sort | uniq
changed:        job@ntt.net 20190127

Is the expected behaviour that force_reload must be set to true to trigger the importer to do something?

[job]

I updated the configuration to the following:

        RPKI:
            authoritative: false
            import_timer: 300
            nrtm_access_list: generic_nrtm_access
            import_serial_source: file:///home/irrd/dbs/RPKI.CURRENTSERIAL
            import_source: file:///home/irrd/dbs/rpki.db

and bumped the CURRENTSERIAL to 187, but it didn't seem to reload the data.

[mxsasha]

This looks like a bug. Do the logs have a message like "Skipping NRTM updates for RPKI, nrtm_host not set." ?

Currently, we run the full import if a serial is not known, or if force_reload is set. In other words, this works, if import_serial_source is not set, and never has been. Instead, this should be based on whether NRTM is configured. With a known serial, it looks like irrd will attempt an NRTM import, which will immediately stop due to lack of a configured NRTM server, and never update.

In general, force_reload should not need to be triggered. It's only purpose is to be able to force a complete reset of database contents, e.g. in case NRTM ran out of sync. In ideal cases, it is never needed. It happens to help here, but won't be needed once we fix the bug.

Quick fix: remove import_serial_source, remove the RPKI row from database_status. That'll have it import the whole file every import_timer. We'll make a proper fix later.

Note that #172 is also still open, which is likely a more suitable method for your use case.