What are possible reasons for RSS replication freezing?
Dennis Melnikov used DELAY_APPLY=7H on an 11.70.FC5XE RSS server nightly, resuming at 5:30 for a data warehouse load, and saw the replica fail to resume or pause up to 30 minutes. Art Kagel said he had not seen it, noted 14.10.FC6 improved RSS, and suggested upgrading and opening an IBM case. Unresolved.
Auto-generated by Claude from the posts below — may be imperfect; read the full thread.
Hi, IBM Informix Dynamic Server Version 11.70.FC5XE Every night we suspend applying logs at RSS with DELAY_APPLY=7H to load consistent data onto data warehouse. At 5:30 we resume replication with DELAY_APPLY=0. Sometimes the replica fails to resume, and restarting the IDS is the only way to revive it. Recently, replication started to pause sometimes for up to 30 minutes and then resumes applying changes without any external intervention. What are possible reasons for such freezes and pauses? ------------------------------ Sincerely, Dennis ------------------------------
Dennis: I would ask why you are delaying apply during the load in the first place? I have not seen this behavior, but then, the sites that I am aware of that use DELAY_APPLY set it permanently on one RSS secondary to use as a fumble-finger recovery site with HDR and/or other RSS running that are not delayed. Art ------------------------------ Art S. Kagel, President and Principal Consultant ASK Database Management Corp. www.askdbmgt.com ------------------------------
Art, We load up data warehouse from RSS. ------------------------------ Sincerely, Dennis ------------------------------
Dennis: Ahh! Make sense. I was looking at it the other way around. So you use DELAY_APPLY to freeze the image in the RSS while you copy data from it to the DW. Interesting, I like it. OK, yeah, you should not be seeing those freezes or pauses. However, note that v11.70 is now three major versions and dozens of patches and bug fixes behind current technology. Particularly, v14.10.FC6 made significant improvements in how RSS and ER work resulting in vastly reduced latency and other bug fixes in between may have already solved your issue. Obviously v11.70 is long out-of-support, so upgrading will be a new purchase, but I think that it may be worth the outlay. Of course my wallet isn't shelling that out, so ... At any rate, while you won't get any fixes unless you have been paying for extended support, it might be worth opening a case with IBM just to determine if this is an old known issue that has been solved. Art ------------------------------ Art S. Kagel, President and Principal Consultant ASK Database Management Corp. www.askdbmgt.com ------------------------------