CE cluster is not starting - LGE and epoch differ

ivan_kovalenko

Hello, everyone. There is something in our installation i cant get right. We've setup CE cluster on vmware, created db, uploaded data. Then shut database down, shut nodes down, upgraded vCPU count, started nodes up. But database is not starting. Nodes are initializing, then shutdown, then lostcontact. On database logs there is this string 2013-08-18 18:13:24.001 Timer Service:0x5843920 [Recover] My local node LGE = 0xa and current epoch = 0xb Is that enought to cancel database startup?

Daniel_Leybovic

Hi! >> Is that enought to cancel database startup? No. LGE can be less than CURRENT EPOCH, typically LGE always less than current epoch. PS: LGE (Last Good Epoch) refers to the most recent epoch that can be recovered.

ivan_kovalenko

Also we have such logs 2013-08-18 18:13:24.001 Timer Service:0x5843920 [Recover] My local node LGE = 0xa and current epoch = 0xb 2013-08-18 18:13:24.128 Spread Client:0x5283080 [Comms] Saw membership message 4352 on Vertica:all 2013-08-18 18:13:24.135 DistCall Dispatch:0x7f841800a100 [Shutdown] Stopping activity on node Does that mean that database is not starting due to difference of LGE (or CPE, which should be current epoch) between cluster nodes?

Daniel_Leybovic

Hi! No. Actually it always true(not "typically") that LGE > CURRENT EPOCH. Please read this: https://my.vertica.com/docs/6.1.x/HTML/index.htm#10734.htm If database startup fails and problem in epochs so Vertica will ask you to start from LGE (but you don't get this message, so problem not in epochs) --- Invalid epoch error: https://my.vertica.com/docs/6.1.x/HTML/index.htm#22387.htm