Red Hat Bugzilla – Bug 83236
rpm leaves behind __db.00* lockfiles after any rpm query
Last modified: 2005-10-31 17:00:50 EST
Description of problem:
After having many troubles with rpm hangs, I performed the rpm
database "recovery" (http://www.rpm.org/hintskinks/repairdb/), which gave no
(I have been trying to install an Openview Operations agent (has rpms), which
was sporadically failing. It uses an install script which basically does
about six "rpm -i" commands to install its rpms.)
Somehow the database is in some state where any rpm command (that actually
reads) will cause it to leave behind the /var/lib/rpm/__db.00* lock files.
After this, database reads will work, but write operations will hang. Even
operations such as "--rebuilddb" will leave behind the lock files. Queries
for non-existant packages will also leave behind the files.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. Get a bad package installation that causes hangs (can be provided)
2. rpm -q anything
3. Note existence of lock files.
4. Delete lock files before you try anything that writes to database...
# ls /var/lib/rpm/__db.00* -l
-rw-r--r--+ 1 root root 8192 Jan 31 10:21 __db.001
-rw-r--r--+ 1 root root 1310720 Jan 31 10:21 __db.002
-rw-r--r--+ 1 root root 360448 Jan 31 10:21 __db.003
No lock files present.
Running base Red Hat 8.0 upgraded from 7.1. rpm version: rpm-4.1-1.06
The __db files are not "lock files" but files that contain locks.
The existence of __db files is not your problem, the existence
of stale locks is.
There are 2 main causes of stale locks.
1) missed SIGCHLD causes rpm-4.1 to "hang"
If using rpm-4.1-1.06, upgrade to rpm-4.1-9 packages
from ftp://people.redhat.com/jbj/test-4.1, the missed
SIGCHLD is fixed there.
2) abnormal termination (e.g. "kill -9") leaves stale locks
It's up to the user to do "rm -f /var/lib/rpm/__db*" in this
A more complete description is in bugzilla, look for the summary
rpm-4.1 hangs: READ THIS FIRST