Add source timestamp to couchbase.Event about go-dcp-kafka HOT 7 CLOSED

Hello @Du-Li, that would be a good feature as far as I know we can get timestamps from CAS value but Couchbase Server should be 7.0 or later. We can add it as a feature maybe any ideas? @erayarslan @mhmtszr

from go-dcp-kafka.

That's awesome. Thanks for your quick attention.

It's a bit disappointing though that this timestamp is only available in CB 7.0 or above. Our current production is still using CB 6.5 and we won't be able to upgrade it in a short term. Is there some way to make it work for CB 6.5?

from go-dcp-kafka.

i checked version 6.5 of the couchbase. there doesn't seem to be a problem with this cas to eventTime conversion.

from go-dcp-kafka.

we implemented in v0.0.37
now you can use with event.EventTime

thank you so much for contribution💯

from go-dcp-kafka.

@erayarslan Thank you for implementing it. I have tested it. However, there was a big gap between this CAS-based timestamp (t2) and the timestamp (t1) I added into the doc when it was generated, actually t2 < t1 by up to a second.

Do you know how CAS was created? What's the physical meaning? I was wondering if it could be rightly used as the timestamp a doc is persisted on CB.

from go-dcp-kafka.

@Du-Li i think its mean "event happen time in couchbase"
so when document create in couchbase, this mutation cas is create time.
or document deleted in couchbase, deletion cas time is delete time.

so when you start consuming for the first time, it is normal to see this gap.
after initial load its done, you will get new fresh events. gap gonna normal.

@oguzyildirim also has idea about DcpTime kinda field to set time(now) when our library consume event.

from go-dcp-kafka.

There is only one edge case about CAS as far as I know. if conflict resolution is to take an older document from the backup it will affect metadata i.e CAS.

I've added dcp latency but not tested yet.

from go-dcp-kafka.