----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/3874/#review5322 -----------------------------------------------------------
bookkeeper-server/src/main/java/org/apache/bookkeeper/bookie/Bookie.java <https://reviews.apache.org/r/3874/#comment11609> I like this interface. It solves the entanglement issue nicely. bookkeeper-server/src/main/java/org/apache/bookkeeper/conf/ServerConfiguration.java <https://reviews.apache.org/r/3874/#comment11604> I still find the meaning of low watermark and high watermark and minor compaction and major compaction confusing. As I understand it, a minor compaction is triggered when low watermark is hit. Major compaction is triggered when the high watermark is hit and no minor compaction ran this round. High watermark must be lower than low watermark. So high is 0.6 for example, and low is 0.9. This is confusing, as one would expect low to be lower than high. Also, if a minor compaction is run on low, it means (in the example above) 90% of entries in the ledger will be copied. For major, 60% will be copied. So minor compaction is heavier than major compaction, which is confusing. I think having a minor and major compaction requires that they run with different regularity, and that major compaction is heavier on the system than the minor compaction. For example, we should only run major compaction once a day, but with a threshold of 0.9, minor compaction can be run every hour with a threshold of 0.6. Both the regularity of the compactions and the threshold should be configurable and explicit. I think we should have: - minorCompactionInterval - minorCompactionThreshold - majorCompactionInterval - majorCompactionThreshold And i think some compaction should be on by default, just to make sure this stuff gets run regularly. - Ivan On 2012-02-20 16:08:51, Sijie Guo wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/3874/ > ----------------------------------------------------------- > > (Updated 2012-02-20 16:08:51) > > > Review request for bookkeeper. > > > Summary > ------- > > bookie server aggregates entries into entry log file. suppose there is lots > of ledgers, each ledger has little messages. so a entry log file would > contains messages from lots of different ledgers. if there is only one ledger > not be deleted, the entry log file would not be removed, whose occupied disk > space could not be reclaimed. > > > This addresses bug BOOKKEEPER-160. > https://issues.apache.org/jira/browse/BOOKKEEPER-160 > > > Diffs > ----- > > bookkeeper-server/conf/bk_server.conf d005d01 > bookkeeper-server/src/main/java/org/apache/bookkeeper/bookie/Bookie.java > d4ece94 > > bookkeeper-server/src/main/java/org/apache/bookkeeper/bookie/EntryLogger.java > aca66e6 > > bookkeeper-server/src/main/java/org/apache/bookkeeper/bookie/GarbageCollectorThread.java > PRE-CREATION > > bookkeeper-server/src/main/java/org/apache/bookkeeper/conf/ServerConfiguration.java > 6bbe943 > > bookkeeper-server/src/test/java/org/apache/bookkeeper/bookie/CompactionTest.java > PRE-CREATION > > bookkeeper-server/src/test/java/org/apache/bookkeeper/bookie/EntryLogTest.java > f661e90 > > Diff: https://reviews.apache.org/r/3874/diff > > > Testing > ------- > > > Thanks, > > Sijie > >
