services/horizon/ledger: Fix memory leak in HistoryDBSource

[bartekn]

PR Checklist

PR Structure

• This PR has reasonably narrow scope (if not, break it down into smaller PRs).
• This PR avoids mixing refactoring changes with feature changes (split into two PRs
  otherwise).
• This PR's title starts with name of package that is most changed in the PR, ex.
  services/friendbot, or all or doc if the changes are broad or impact many
  packages.

Thoroughness

• This PR adds tests for the most critical parts of the new functionality or fixes.
• I've updated any docs (developer docs, .md
  files, etc... affected by this change). Take a look in the docs folder for a given service,
  like this one.

Release planning

• I've updated the relevant CHANGELOG (here for Horizon) if
  needed with deprecations, added features, breaking changes, and DB schema changes.
• I've decided if this PR requires a new major/minor version according to
  semver, or if it's mainly a patch change. The PR is targeted at the next
  release branch if it's not a patch change.

What

This commit fixes a memory leak in HistoryDBSource.NextLedger method. The method is usually used like in stream_handler.go:

go/services/horizon/internal/render/sse/stream_handler.go

Lines 73 to 79 in 00d7d2f

	select {
	case currentLedgerSequence = <-handler.LedgerSource.NextLedger(currentLedgerSequence):
	continue
	case <-ctx.Done():
	stream.Done()
	return
	}

When streaming connection is closed but the are no new ledgers (ex. due to database or network issues) the goroutines started in HistoryDBSource.NextLedger will not be closed.

To fix this ledger.Source interface was updated with Close() method that should be called when a ledger source is no longer used and it exist internal goroutine. The code was refactored to create a new HistoryDBSource for each streaming request (so it can be closed when handler execution ends).

Why

When there are no new ledgers, the goroutines started in HistoryDBSource.NextLedger will leak and will eventually consume all available memory, like in #2535.

Known limitations

The new code creates a new HistoryDBSource for each streaming request (instead of reusing a single HistoryDBSource in all streaming requests). This should have a really minimal impact on memory usage.

[tamirms]

another go routine may write to source.closed so I think technically the behavior is undefined. I think it would be a good idea to use a lock or use a context to signal cancelation

[bartekn]

You're right. I was under the impression that single read/single write is totally fine but https://golang.org/ref/mem is clear that it's not. Fixed by using explicit synchronization.