Replication on "Completing Display procedure of Procedure"

[ChrisF79]

I manually replicate a cube from a remote location to our main TM1 server. The particular cube and dimensions takes me around 15 minutes to do manually. I set the cube to sync and built a chore that replicates just that one cube but it takes 40 minutes now. One other thing that's strange is that in Perspectives, I can watch it replicate normally and all looks good until it gets to "Completing Display procedure of Process" in the progress box and just sits there for the remainder of the 40 minutes. Any idea why that may be?

[ChrisF79]

Bump?

[lotsaram]

Have you looked in TM1 top while doing the replication? Depending on the size of the cube the commit phase could conceivably take quite a while. Also check in both top and the message log if there are any rollbacks.

[Steve Vincent]

Never seen that message before but i've seen enough issues with replication not to trust it as far as i can throw a Chieftain Tank. Its very sensitive to log files and their size especially, have they increased significantly on either server?

[ChrisF79]

TM1Top just shows it working. We haven't seen any big changes in the log files.

As for not trusting replications, I agree that they're pretty finicky. However, if you have servers in multiple locations, what do you do to move data around?

[Steve Vincent]

i wrote my own set of TIs to try and do the job instead. its not perfect & took a lot of time to figure out but it is customised to our needs and flexible, unlike replication which won't even let you edit an existing setup. now that it doesn't rely on log files it takes a predictable amount of time, always works and avoids some of the bugs i found in 9.5.2's implimentation of replication. i don't have merging of data issues or servers on a WAN to contend with tho, all 4 machines are in the same datacentre and the data copied is always like for like, star to planets with one way data traffic only.

[tomok]

unlike replication which won't even let you edit an existing setup.

Unless they've changed the synch/rep architecture in the latest few versions that's not entirely true. Most of the replication switches are set in various system cubes and can be modified by simply changing the values in some of those cells. I used to have a replication scheme in an earlier life and I was able to drastically alter the replication setup by just changing those values. Things like whether or not you replicate dimensions, rules, just data, etc. can all be changed after you do the setup. Granted there is no documentation on this and I figured it all out by trial and error, but it can be done.

[Steve Vincent]

Except i tried that in 9.5.2 and it continued to do things i didn't want it to. Its so sparsly documented i'm not prepared to trust it anymore. Also, failing to replicate attribute values when in a chain of servers because the log files are missing correct time stamps was one bug too many, hence doing it myself