On Wed, Aug 16, 2017 at 10:18 PM, Guanghao Zhang <[email protected]> wrote:
> About compatibility, there is one incompatible change about the replication > TableCFs' config. The old config is a string and it concatenate the list of > tables and column families in format "table1:cf1,cf2;table2:cfA,cfB" in > zookeeper for table-cf to replication peer mapping. When parse the config, > it use ":" to split the string. If table name includes namespace, it will > be wrong (See HBASE-11386 > <https://issues.apache.org/jira/browse/HBASE-11386>). It is a problem > since > we support namespace (0.98). So HBASE-11393 > <https://issues.apache.org/jira/browse/HBASE-11393> (and HBASE-16653 > <https://issues.apache.org/jira/browse/HBASE-16653>) changed it to a PB > object. When rolling update cluster, you need rolling master first. And the > master will try to translate the string config to a PB object. But there > are two problems. > 1. Permission problem. The replication client can write the zookeeper > directly. So the znode may have different owner. And master may don't have > the write permission for the znode. It maybe failed to translate old > table-cfs string to new PB Object. See HBASE-16938 > <https://issues.apache.org/jira/browse/HBASE-16938> > 2. We usually keep compatibility between old client and new server. But the > old replication client may write a string config to znode directly. Then > the new server can't parse them. > > PS: HBASE-11386 is a problem since 0.98. But there are not too many users > report this problem. So it maybe not a big incompatible issue...... > > Excellent writeup Guanghao. File a blocker with the above. We need to have some facility to get folks over the above hump. Thanks, St.Ack > 2017-08-15 1:54 GMT+08:00 Stack <[email protected]>: > > > Heads-up: > > > > I'm about to put up an hbase-2.0.0-alpha2 Release Candidate. Theme is > > updated dependencies, reliance on relocated popular libs (guava, netty, > > protobuf), purge of checked-in generated src, and > master-carries-no-regions > > by default. > > > > alpha3 I hope will follow soon after (end-of-August?). Its theme will be > > settling the APIs and compatibility (At first blush, we are not looking > too > > bad; our Sean ran some tests over weekend that have hbase-1 client > running > > against an hbase-2 cluster....). The Coprocessor Interface revamp should > be > > done by alpha3 (i.e. returning Interfaces rather than Implementations, > and > > our shutdown of CPs accessing classes in hbase marked InterfaceAudience). > > We'll also have purged thirdparty classes from our API; e.g. guava 0.12 > > Service showing through in our replication API and protobufs in Admin > > Interface. On alpha3, we will have to do a bunch of outreach to make sure > > our downstreamers are up on what is coming down the pipe. > > > > Beta1 in mid-September? > > > > I encourage you to check out the items marked for hbase2: > > https://issues.apache.org/jira/projects/HBASE/versions/12327188 Edit as > > you > > see appropriate. Punt if you know the JIRA will not get any attention in > > next month or so. > > > > A bunch of issues marked blocker are unassigned. I'll leave them as is > > another while but I'll boot them soon. > > > > While I have your attention: > > > > + I think we should leave thrift version at 0.9.3. Moving hbase thrift to > > 0.10.0 will break existing clients. The change is easy enough if folks > need > > to upgrade their hbase thrift. See HBASE-18591. > > + Upgrade from 0.94 is disallowed. You have to get to 1.0 first (0.98?). > > > > St.Ack > > > > > > > > On Wed, Aug 2, 2017 at 9:43 AM, Stack <[email protected]> wrote: > > > > > > > > > > > On Tue, Aug 1, 2017 at 2:06 PM, Josh Elser <[email protected]> wrote: > > > > > >> > > >> > > >> On 7/31/17 9:00 AM, Stack wrote: > > >> > > >>> On Mon, Jul 24, 2017 at 12:25 PM, Josh Elser<[email protected]> > > wrote: > > >>> > > >>> ... > > >>>> > > >>>> I like the idea of this also hitting 2.0 as it would make the > feature > > a > > >>>> bit more "real", but am obviously a little nervous (I have no reason > > to > > >>>> be > > >>>> nervous though). I am pretty happy with the feature in terms of how > > >>>> much it > > >>>> is covered via testing. > > >>>> > > >>>> https://issues.apache.org/jira/browse/HBASE-17748 > > >>>> > > >>>> > > >>>> > > >>>> Sounds good to me. Whats involved? Backport? If so, +1 Josh. > > >>> > > >>> Last think on space quota says that need doc too. See 'Space Quota' > in > > >>> here: > > >>> https://docs.google.com/document/d/1WCsVlnHjJeKUcl7wHwqb4z9i > > >>> Eu_ktczrlKHK8N4SZzs/edit#heading=h.wuw3a6jukzo5 > > >>> Does this little section need an update Josh? > > >>> > > >>> Thanks, > > >>> S > > >>> > > >> > > >> Yep, just a couple of cherry-picks. Good test coverage and some docs > > >> already included for 17748. Happy to put that on my plate if you're > > good > > >> with it. I can reasonably assume that no one is against it :) > > >> > > >> I think I had knocked out docs for the "phase 1" stuff before we > merged > > >> it in from the original feature branch. I'll double check and update > the > > >> gdoc. Perhaps this was just a timing thing. > > >> > > > > > > Thanks Josh, > > > S > > > > > >
