Bug 6814

Summary:	rpm log file size stays the same no matter what
Product:	[Build System, Metadata & Runtime] BSPs	Reporter:	Costin <costin.c.constantin>
Component:	bsps-configuration	Assignee:	Mark Hatle <mark.hatle>
Status:	RESOLVED NOTABUG	QA Contact:	Costin <costin.c.constantin>
Severity:	normal
Priority:	Medium	CC:	mark.hatle, yp.bsp.watcher, yp.watcher
Version:	1.7
Target Milestone:	1.7
Hardware:	x86
OS:	x86_64
Whiteboard:
OS type for building Yocto:	---	Type of Regression:	---
Verified:		Documentation change:	Don't know

Description Costin 2014-10-07 12:41:05 UTC

Hello,

Found this bug while testing genericx86-64 on Sugarbay(5367855fe56db0c09b8fa7bb423b9de18d11eecc). 
The bug is related to TC195 "Check rpm install/removal log file size". TC content:
1. After system is up, check the log file size after rpm/smart/zypper install/removal
2. For rpm, there will be some database files under /var/lib/rpm/, named as "__db.xxx" and there will be some log files under /var/lib/rpm/log, named as "log.xxxxxx". Each file will occupy about 10MB.
3. After several rpm/smart/zypper install/removal, rpm will create several log files under /var/lib/rpm/log, which eat lots of system disk space.

No matter how many packages I install/remove, the /var/lib/rpm/log/log.* file stays the same in size (meaning no more than one log file is present). Interesting though, each time I install/remove something, the log file changes it's time stamp.

For an indepth test I made this small script (where pkgs to install are on the same path with the script):

#!/bin/bash

for i in `seq 100` 
do 
  rpm -ivh groff*
  rpm -ivh man*
  rpm -e man
  rpm -e groff
done

Regards,

Costin

Comment 1 Costin 2014-10-07 13:48:37 UTC

hello,

Just performed same test case on genericx86 on AtomPC and results are similar. Same hash used. More interesting is the fact that the log size is the same as on genericx86-64, 10485760 bytes

Costin

Comment 2 Mark Hatle 2014-10-09 15:31:23 UTC

The 'log' file is the database journal.  Once the journal fills or on various recovery or other operations, the journal is dumped into the data itself.

So the log simply needs to be large enough to handle a single transaction.

There may be specific settings to lower the size of the log file, look at the BerkleyDB manual and the DB_CONFIG file in /var/lib/rpm.