Yah, actually I would be interested in such experiment, of course not on our client data like last time :) but we have two postgres clusters with slon in our lab, so we can use it freely.. I am not very strong in slons inside structure, and still thinking why it needs this timeout? Why it is not possible to divide all events in parts which can be replicated during one timeout? and in second it could replicate second part?
Lukas > Lukas wrote: >> Hello, >> >> thank you Christopher for very professional answer, I believe it was >> right solution, but to late, I recreated node and not it is working.. >> >> Why slon tries to make all SYNCs in one timeout, isn't it better to >> make >> them one by one? >> >> > The thing is, that query is a couple steps indirect from *doing* the > replication work; the query that is timing out is the one that, loosely > speaking, is asking: > > "What work (e.g. - set of events) is outstanding for me to work on?" > > Jan and I just had a chat about this on IRC; the concern is that if this > query limits the result set, it's not correctly assessing "what work is > outstanding?" > > I'm interested in doing an experiment where we put a LIMIT on that > query, whether as actual experiment, or, better, analytically as a > Gedankenexperiment, but that is definitely not something I'd be keen on > using to experiment on your data :-). > > What we can and will do is to change the 300 second timeout time into a > parameter that may be configured at runtime (e.g. - in a slon.conf > file). That won't help you today, as you've already dropped the node, > but it can help others later... > > -- > This message has been scanned for viruses and > dangerous content, and is believed to be clean. > > -- This message has been scanned for viruses and dangerous content, and is believed to be clean. _______________________________________________ Slony1-general mailing list [email protected] http://gborg.postgresql.org/mailman/listinfo/slony1-general
