Node recovery never ends

Jack_Kidwell

2013-04-26 00:45:07.549 Timer Service:0x2aaba5d0b5e0-e0000004ad52ac [Recover] Get Local Node LGE: node current cpe is 69f497c 2013-04-26 00:45:46.226 Timer Service:0x2aaba5d0b5e0-e0000004ad52ad [Recover] Get Local Node LGE: node current cpe is 69f497c 2013-04-26 00:45:46.347 Timer Service:0x2aaba5d0b5e0-e0000004ad52ae [Recover] Get Local Node LGE: node current cpe is 69f497c 2013-04-26 00:46:09.127 Timer Service:0x2aaba5d0b5e0-e0000004ad52af [Recover] Get Local Node LGE: node current cpe is 69f497c 2013-04-26 00:47:05.448 Timer Service:0x2aaba5d0b5e0-e0000004ad52b0 [Recover] Get Local Node LGE: node current cpe is 69f497c ... 2013-04-26 01:05:08.775 Timer Service:0x2aaba5d0b5e0-e0000004ad52e9 [Recover] Get Local Node LGE: node current cpe is 69f497c 2013-04-26 01:05:08.891 Timer Service:0x2aaba5d0b5e0-e0000004ad52ea [Recover] Get Local Node LGE: node current cpe is 69f497c 2013-04-26 01:05:28.054 Timer Service:0x2aaba5d0b5e0-e0000004ad52eb [Recover] Get Local Node LGE: node current cpe is 69f497c 2013-04-26 01:05:28.167 Timer Service:0x2aaba5d0b5e0-e0000004ad52ec [Recover] Get Local Node LGE: node current cpe is 69f497c

Jack_Kidwell

Vertica 6.1.1 Is there something to look for in logs that provide insight as to how to resolve?

Jack_Kidwell

After several hours, the LGE has advanced. Perhaps I am impatient.

[Deleted User]

Hi Jack, thanks for reporting your troubles. We're glad you got it working but we'll look into this to see if the timing you experienced is typical.

colin_loghin

Documentation states: Scenario: Tuning for Recovery Scenario You have a large database that contains a single large table with two projections, and with default settings, recovery is taking too long. You want to give recovery more memory to improve speed. Solution Set the PLANNEDCONCURRENCY and MAXCONCURRENCY setting of the recovery pool to 1 so that recovery can take as much memory as possible from the GENERAL pool and run only one thread at once. Note: This setting could slow down other queries in your system. =============== I would be interested in how to tune recovery in a real world system with thousands of tables and maybe tens of tables to incrementally recover. What is the proper degree of parallelism and what are the best practices in adjusting DML operations ( insert/ update/delete) to minimize recovery times.