See http://lists.slony.info/pipermail/slony1-bugs/2010-May/000495.html Basically no sync events (other than when slon starts) are being generated on nodes that are not origins for any set. This makes the sl_status table seem out of date. This change was checked in on Mar 14 2007. Current thinking is that this is actually a bug.
We noticed this behavior change when moving from 1.2x to 2.0x as it made some health monitors conducting sl_status lag-time polling uninformative on non-origin nodes. In the archives, I notice at least one other person encountering the problem too http://permalink.gmane.org/gmane.comp.db.postgresql.slony1.general/9029 This bug is marked as applying to v1.2, but the behavior effects v2.0 as well and gives non-origin nodes the appearance of accumulating tremendous replication lag.
Created attachment 49 [details] patch against 2.0 to restore sync generating on all nodes This patch reverts the checkin (updated to the 2.0 stream) the original commit that disabled generating sync messages on leaf nodes.
(In reply to comment #2) > Created an attachment (id=49) [details] > patch against 2.0 to restore sync generating on all nodes > > This patch reverts the checkin (updated to the 2.0 stream) the original commit > that disabled generating sync messages on leaf nodes. The code is a bit more complex than I'd have expected, but this looks reasonable to me. (I guess it was more complex to suppress the events than I'd expected!)
Committed to master and REL_2_0_STABLE fe2c2040fcc454fef9c4da85a518c08840b49c2f 572581c4d80fd95206aed1910fb22d7119aa4be4 Note this has been left as is on the 1.2 branch