Thanks Pankaj,
I had not the time to read all your message, very useful I guess.
I will post all this (with Deyan's remark too) on Wiki after having done my own
opinion...
Jacques
From: <[email protected]>
Hi Deyan n others
I have also worked & solved many sync problems in ofbiz in my application:
1) As you said while sync is in progress if server crashes, the status in
EntitySync table gets in running state and we can't sync further until we
manually change it even if we restart server it remains in running state
To solve this problem: I created a new entity SyncStatus on client side
to maintain the status of sync, in that ist status is Not Started, when
service runEntitySync or runPullEntitySync starts I set the status of
field as running and when service returns (whether fail or successful) I
updated status to finish.
I also added code, that if server restarts and if this field status is
running change it to not started, By this I solved the above problem:
*) If client crashes, then after restart, status in SyncStatus table will
be Not Started.
*) If server crashes, then service will be failed and status will be
finished and no further problem.
2) There were other problems like if sync(push or pull) is running, and
one other job of same service is also created then we get exception "An
instance is already running" By using above status in SyncStatus table
when A job of sync is running, I just cancelled out all other jobs of same
type.
3) I also solved "Connection refused" problem by same method even I
displayed message on screen rather than giving exception on console.
4) I did not get your point of updating time stamp, I did not get this
problem because In push or pull sync the successful timestamp is updated
on EntitySync table on client or MCS side only if sync is successful, if
server crashes or sync fails this field is not updated, but yes there are
some issues:
*) If we have 2 hours of sync, n sync fails on last minute. then in next
sync as sync successful status has not been revisesd, next sync will pick
the status of previous one, as a solution I think we need to keep updating
syn status in EntitySync table after particular intervals. but how to
decide these intervals b'coz tables are linked with foreign key
constriants.
*) One problem with connection refused in push and pull case is that, In
pull sync first connection is established then data is prepared for sync
while in push sync first data is prepared and then conncetion is
established. So in pull sync we get connection refused problem in earlier
stage than push sync. We can say in push sync if there is no MCS (Or MCS
IP address is wrong), Our code is doing efforts in preparing data that
effort is of no use.
5) If we place order on POS side and try to perform push sync, some time I
was getting random exceptions due to order header or Order Status table, I
solved the problem.
Explain: In placing order (or others) if you have used a database trigger
(eecas) like if an entry is created on A table create an entry on B table.
Now wile sync if MCS is also having same trigger then as a entry is
created (by sync) on A table , trigger is fired and an entry for B is
created. Now for one A, B will be having two entries that creates random
exception
So I want to ask, that mean for sync there should not be ant trigger on
opposite side(means for push MCS and for pull POS), how we will maintain
that ???
I did not get your concept of using XML file inspite of RMI. Could you
please give some details regarding that???? Even I agree with the fact
that RMI takes long time.
Regards:
Pankaj Jain
______________________________________________________________________
