[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: stuck transactions?



Peter Mogensen wrote:
Hi,

I notice that when I run db4.8tat -t the Active Transaction list does
not change very often. I would have expected it to not be the same
twice. So I guess I have a lot of stuck transactions?

No.

I use back_hdb and slapd is under some load.

Could this be related to the messages in the log:

No.

Dec 15 15:09:19 server1 slapd[32060]: =>  bdb_idl_delete_key: c_del id
failed: DB_LOCK_DEADLOCK: Locker killed to resolve a deadlock (-30994)
Dec 15 15:09:19 server1 slapd[32060]: conn=40803 op=18: attribute
"entryCSN" index delete failure

Active transactions:
         80000003: running; pid/thread 32060/47089260848688; begin LSN:
file/offset 0/0
         80000004: running; pid/thread 32060/47089260848688; begin LSN:
file/offset 0/0
         80000006: running; pid/thread 32060/1127287120; begin LSN:
file/offset 0/0
         80000007: running; pid/thread 32060/1110501712; begin LSN:

Notice that these TXNs all have empty file/offset, which means they have not generated any modifications. They are the long-lived read-only TXNs that back-bdb/hdb use for internal purposes. Completely normal.

--
  -- Howard Chu
  CTO, Symas Corp.           http://www.symas.com
  Director, Highland Sun     http://highlandsun.com/hyc/
  Chief Architect, OpenLDAP  http://www.openldap.org/project/