Update watch delay increment to 50ms

[MikeGoldsmith]

Updated the increment amount to 50ms for watch index retry delay.

The sentence indicates starting at 50ms with an upper bound of 1000ms but originally had the increment set at 500ms, which would only allow 2 increments before the upper bound was reached.

[daschl]

@MikeGoldsmith looks like a good improvement, I think we need full +1 from all owners so we can merge this (double check with the impls and if not conform file tickets).

edit: I just checked, and java indeed uses 500 ->

private static final Delay INDEX_WATCH_DELAY = Delay.linear(TimeUnit.MILLISECONDS, 1000, 50, 500);

ping @couchbaselabs/sdk-team

[mnunberg]

Python:

    def n1ql_index_watch(self, indexes,
                         timeout=30, interval=1, watch_primary=False):

Python uses a fixed single-second interval, but this is easy to adjust by the user at a per-operation level. I haven't yet got around to implementing a lower/upper bound.

[MikeGoldsmith]

.NET currently uses a fixed 50ms interval

private static readonly TimeSpan WatchIndexSleepDuration = TimeSpan.FromMilliseconds(50);

[daschl]

@couchbaselabs/sdk-team heads up! 🔥

• Java
• .NET
• Node
• PHP
• Python
• Go
• C

[daschl]

@MikeGoldsmith the only thing that might make sense is in additon to add an explicit "errata" to the rfc at the end which mentions this change.

[mnunberg]

While this might be out-of-scope for this RFC, what's the recommended retry interval for clients which don't use a linearly increasing interval?

[MikeGoldsmith]

@daschl @mnunberg I've added a proposed static delay of 200ms if an increasing delay is not possible and added the errata section to describe changes

[daschl]

250/200ms?

[MikeGoldsmith]

@daschl I believe this is ready for review.