Drop All Failure?

I’ve been having problems with “drop all” when I want to drop the whole database and start over. The web request times out … or gets reset.

curl -X POST -H "X-Dgraph-CommitNow: true" --data '{"drop_all": true}' http://...:8081/alter
curl: (56) Recv failure: Connection reset by peer

Afterwards I see a few of the predicates remaining in the database. If I run the drop again, those predicates disappear. The second request also seems to just hang for a very long time and not return.

The only information in the logs is for zero:

2018/03/27 20:12:39 oracle.go:85: purging below ts:62871, len(o.commits):0, len(o.aborts):0, len(o.rowCommit):0
2018/03/27 20:12:59 oracle.go:85: purging below ts:62872, len(o.commits):0, len(o.aborts):0, len(o.rowCommit):100:

I’m not sure that is related to the second drop request.

Look at /debug/vars, I see:

"dgraph_predicate_stats": {"Transfer": 229285, "amount": 229271, "at": 229285, "from": 229270, "i.at": 229285, "i.from": 229270, "i.to": 229285, "net": 9895, "to": 229285},

If I restart each of the docker containers for the servers (3 of them), the “stats” reset:

"dgraph_predicate_stats": {},

I’m not quite sure where to look from an administration perspective to see the status of the drop or size of various predicates.

…and the timeout seems to be relative to the amount of data loaded. If I start with an empty database, load a small amount of data, the drop all succeeds. If I load a lot of data, the drop all takes a long time and eventually seems to timeout or get reset (which might just be the connecting getting dropped because of networking).

Is there a way to get a certificate for a long-running transaction such as “drop all”.

This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.