Add additional logs to show progress of reindexing operation.

[martinmr]

The logs are V(1) like the rest of the reindexing logs.

Fixes #3741

---

This change is 

[gitlw]

Reviewable status: 0 of 1 files reviewed, all discussions resolved (waiting on @manishrjain)

[danielmai]

Logs are a bit too noisy. I tried doing the following:

1. Load up 21-million movie data set.

2. Reindex the name predicate by adding the exact index:

   curl localhost:8180/alter -d 'name: string @index(hash, term, exact, fulltext, trigram) @lang .'
   

I saw over 2000 lines of log output about the reindexing progress in the alpha logs over the timespan of 1m23s.

Maybe decrease how often we print out the logs (say, every 100k keys) or output a log every so often e.g., every 5 seconds.

Reviewable status: 0 of 1 files reviewed, all discussions resolved (waiting on @manishrjain)

[danielmai]

See attached alpha1.log for the progress logs when reindexing name in 21-million movie data set.

[danielmai]

I also already see logs like these that mention the index rebuilding process and time elapsed. This comes from posting/index.go:

Rebuilding index for predicate name Time elapsed: 37s, bytes sent: 0 B, speed: 0 B/sec

Reviewable status: 0 of 1 files reviewed, all discussions resolved (waiting on @manishrjain)

[manishrjain]

Please do not merge this as it is. See my comments. Also, @danielmai 's comments.

Reviewable status: 0 of 1 files reviewed, 1 unresolved discussion (waiting on @manishrjain and @martinmr)

---

posting/index.go, line 531 at r1 (raw file):

		}

		counterChan <- struct{}{}

Pushing on a channel for every key is very contentious. This would single-handedly kill performance.

[martinmr]

I removed the logs during the stream phase since the stream framework is already providing logs and there's no way to do this without having to use some kind of contention to accurately count the number of keys processed.

I added logs for the phase that writes the deltas to disk. I also updated the logs to all have the same prefix ("Rebuilding index for predicate ...").

Reviewable status: 0 of 1 files reviewed, 1 unresolved discussion (waiting on @manishrjain)

---

posting/index.go, line 531 at r1 (raw file):

Previously, manishrjain (Manish R Jain) wrote…

Pushing on a channel for every key is very contentious. This would single-handedly kill performance.

Removed this code.

addressed comments

[gitlw]

Reviewable status: 0 of 1 files reviewed, 1 unresolved discussion (waiting on @manishrjain)

[manishrjain]

Reviewed 1 of 1 files at r2.
Reviewable status: complete! all files reviewed, all discussions resolved

[manishrjain]

Reviewable status: complete! all files reviewed, all discussions resolved