[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: Antw: Re: Partial SyncRepl replication



On Tue, 24 May 2016 at 10:07am, Ulrich Windl wrote:

Frank Swasey <Frank.Swasey@uvm.edu> schrieb am 24.05.2016 um 14:10 in Nachricht
<alpine.OSX.2.20.1605240743160.16039@vc0ns51415.vag.hiz.rqh>:
On Thu, 19 May 2016 at 3:10pm, Howard Chu wrote:

The contextCSN is used in replication, but it does not get replicated
itself.
Each replica maintains its own.

I was testing my partial replication against my test MMR setup.  I have
discovered that the database suffix in the two MMR's have two contextCSN
values.  The code I was using to test was checking the first contextCSN
value returned from the "master" server - which is not the "current" CSN
against the value generated by the replica.  I have not (yet) figured
out what that first of the two values is indicating.

I think there is one CSN per ServerID, like this:
contextCSN: 20130722065709.189194Z#000000#000#000000
contextCSN: 20160222141054.581517Z#000000#001#000000
(or another context)
contextCSN: 20130719093756.074776Z#000000#000#000000
contextCSN: 20160517103616.692516Z#000000#001#000000
contextCSN: 20151007103227.123574Z#000000#002#000000


It appears that each master (I have two, ServerID 1 and 2 respectively) sets its own contextCSN (the 00[012] segments are the serverid numbers) and I guess updates based on entryCSN values for the other servers. My own two servers have the following contextCSN values right now:

(ServerID 1)
contextCSN: 20160525000023.928191Z#000000#000#000000
contextCSN: 20160525135840.176813Z#000000#001#000000
contextCSN: 20160525102003.058340Z#000000#002#000000

(ServerID 2)
contextCSN: 20160525000023.928191Z#000000#000#000000 contextCSN: 20160525135840.176813Z#000000#001#000000
contextCSN: 20160525135906.347736Z#000000#002#000000

(Replica)
contextCSN: 20160525102935.619902Z#000000#000#000000 contextCSN: 20160525135840.176813Z#000000#001#000000
contextCSN: 20160525135906.347736Z#000000#002#000000

So, it appears they have gone split brain on me - since they have not gotten to a place where they agree on what ServerID2's contextCSN should be.

The ServerID 0 records are the result of performing the initial slapadd from the non-MMR production system.


--
Frank Swasey                    | http://www.uvm.edu/~fcs
Sr Systems Administrator        | Always remember: You are UNIQUE,
University of Vermont           |    just like everyone else.
  "I am not young enough to know everything." - Oscar Wilde (1854-1900)