> On Nov. 18, 2016, 11:59 p.m., Vadim Spector wrote: > > Since executeTransactionWithRetry() calls executeTransaction(), and both > > APIs are public, I wonder if we care about separate counters for both, to > > distinguish between intermittent and more serious failures. > > > > For the client code, executeTransactionWithRetry() is successful even if it > > some calls to executeTransaction() failed. With two separate counts, if we > > take the number of failed calls to executeTransactionWithRetry() multiplied > > by _transactionRetryMax_, and subtract it from the number of failed > > executeTransaction() calls, we get the number of failed _direct_ calls to > > executeTransaction(), which gives us a complete picture. > > > > Just a thought, not sure if it really matters.
I don't think we need to distinguish at this level - I do want to see failures grouped by exception type and this is captured here. In general these counters are useful as a way to get a feel for a system health, so a simpler model is better. - Alexander ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/53905/#review156357 ----------------------------------------------------------- On Nov. 18, 2016, 10:50 p.m., Alexander Kolbasov wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/53905/ > ----------------------------------------------------------- > > (Updated Nov. 18, 2016, 10:50 p.m.) > > > Review request for sentry, Colm O hEigeartaigh, Colin Ma, Hao Hao, Sravya > Tirukkovalur, and Vadim Spector. > > > Repository: sentry > > > Description > ------- > > SENTRY-1518: Add metrics for SentryStore transactions > > > Diffs > ----- > > > sentry-service/sentry-service-server/src/main/java/org/apache/sentry/provider/db/service/persistent/TransactionManager.java > 3d0a6ab0b514cb9fbead4b8e527f7cb84865ce28 > > Diff: https://reviews.apache.org/r/53905/diff/ > > > Testing > ------- > > > Thanks, > > Alexander Kolbasov > >
