Thanks. I missed it grepping. Each hfile has written into its metadata its maximums sequence (edit) number -- the sequence id of its last entry. MAX_SEQ_ID_KEY is the key used writing this attribute into the hfile metadata map.
St.Ack On Tue, Aug 18, 2009 at 9:55 AM, chunping rao <[email protected]> wrote: > hi: > stack > MAX_SEQ_ID_KEY is in the StoreFile.java of package > org.apache.hadoop.hbase.regionserver.....Sorry ,i do not known exactly the > version of my hbase. > > > 2009/8/19 stack <[email protected]> > > > Oops. Sorry. Answered wrong question. > > > > I grepped source in 0.19 and 0.20 branches and don't see MAX_SEQ_ID_KEY. > > Where are you seeing that? > > > > Thanks, > > St.Ack > > > > > > > > On Tue, Aug 18, 2009 at 9:43 AM, stack <[email protected]> wrote: > > > > > On each edit, a sequence id is incremented. When edits are applied to > > the > > > commit log their sequence id is noted along with the entry. When we > > flush, > > > the sequence id of the flushed edits are noted in the flush/store file. > > The > > > regionserver will let go of commit logs (hlogs in hbase parlance) if > all > > its > > > edits have been flushed to the filesystem. It uses the sequence id > > figuring > > > when it can do this. Sequence id is also used figuring which of a set > of > > > flush/store files is the newer (the file denoted by the larger sequence > > id > > > will be newer) among other things. > > > > > > Does this help? > > > > > > St.Ack > > > > > > > > > On Tue, Aug 18, 2009 at 3:18 AM, chunping rao <[email protected]> > > wrote: > > > > > >> hi: > > >> all > > >> in the source code of hbase-0.20.0,i can't understand the use of > the > > >> attribute "MAX_SED_ID_KEY" ,what is the role of "MAX_SEQ_ID_KEY"? > > >> > > > > > > > > >
