[9fans] replica question

[Rudolf Sykora]

Hello,

Yesterday I read replica(8), played with it a bit, and I have the
following question:

Say I have a directory A which I—from time to time—want to replicate
to directory B. With replica/updatedb I maintain two files, A.db and
A.log, describing the situation in the A directory.

If A.log has existed since the time replica/updatedb was run for the
first time, I may use this log as an input to replica/applylog and
create a replica of A in B.

However, I don't understand what one should do if this log was removed
or somehow shortened (so that it does not mention the files' addition
anymore). Some such shortening is, I believe, reasonable, since after
say 10 years nobody is probably interested in all the files' changes.
But, if I understand, it is exclusively the log itself (i.e., not the
server database file A.db) that is used for synchronization by
replica...

Is there any more thorough explanation somewhere describing how replica works?

Thank you for hints/explanation!

Ruda

[erik quanstrom]

Post by Rudolf Sykora
However, I don't understand what one should do if this log was removed
or somehow shortened (so that it does not mention the files' addition
anymore). Some such shortening is, I believe, reasonable, since after
say 10 years nobody is probably interested in all the files' changes.
But, if I understand, it is exclusively the log itself (i.e., not the
server database file A.db) that is used for synchronization by
replica...

read about compactdb, which is in replica(8).

- erik

[Rudolf Sykora]

Post by erik quanstrom
read about compactdb, which is in replica(8).

Compactdb works on the .db (simplifies it) but does nothing to the .log.
Applylog only reads the .log.
I don't see any way how compactdb would help here.

Thanks
Ruda

[erik quanstrom]

Post by Rudolf Sykora

Post by erik quanstrom
read about compactdb, which is in replica(8).

Compactdb works on the .db (simplifies it) but does nothing to the .log.
Applylog only reads the .log.
I don't see any way how compactdb would help here.

the way i've always done this is to copy the db and then
generate the log i wanted that represented the differences
i was interested in. i did to the extent of replicating the
/n/dump from a 32-bit file server onto a 64-bit file server.

(geoff did a great job on that. not a single problem with
the 64-bit file server stuff in 7 years.)

- erik