6

So.. I've installed Logstash, and instead of using the logstash shipper (because it needs the JVM and is generally massive), I'm using rsyslogd with the following configuration.

# Use traditional timestamp format
$ActionFileDefaultTemplate RSYSLOG_TraditionalFileFormat
$IncludeConfig /etc/rsyslog.d/*.conf


# Provides kernel logging support (previously done by rklogd)
$ModLoad imklog
# Provides support for local system logging (e.g. via logger command)
$ModLoad imuxsock

# Log all kernel messages to the console.
# Logging much else clutters up the screen.
#kern.*                                                 /dev/console

# Log anything (except mail) of level info or higher.
# Don't log private authentication messages!
*.info;mail.none;authpriv.none;cron.none;local6.none            /var/log/messages

# The authpriv file has restricted access.
authpriv.*                                              /var/log/secure

# Log all the mail messages in one place.
mail.*                                                  -/var/log/maillog


# Log cron stuff
cron.*                                                  /var/log/cron

# Everybody gets emergency messages
*.emerg                                                 *

# Save news errors of level crit and higher in a special file.
uucp,news.crit                                          /var/log/spooler

# Save boot messages also to boot.log
local7.*                                                /var/log/boot.log

In /etc/rsyslog.d/logstash.conf there are 28 file monitor blocks using imfile

$ModLoad imfile   # Load the imfile input module
$ModLoad imklog   # for reading kernel log messages
$ModLoad imuxsock # for reading local syslog messages

$InputFileName /var/log/rabbitmq/startup_err
$InputFileTag rmq-err:
$InputFileStateFile state-rmq-err
$InputFileFacility local6
$InputRunFileMonitor
....
$InputFileName /var/log/some.other.custom.log
$InputFileTag cust-log:
$InputFileStateFile state-cust-log
$InputFileFacility local6
$InputRunFileMonitor
....
*.* @@10.90.0.110:5514

There are 28 InputFileMonitor blocks, each monitoring a different custom application logfile..

If I run

[root@secret-gm02 ~]# lsof|grep rsyslog
rsyslogd   5380        root  cwd       DIR              253,0       4096          2 /
rsyslogd   5380        root  rtd       DIR              253,0       4096          2 /
rsyslogd   5380        root  txt       REG              253,0     278976    1015955 /sbin/rsyslogd
rsyslogd   5380        root  mem       REG              253,0      58400    1868123 /lib64/libgcc_s-4.1.2-20080825.so.1
rsyslogd   5380        root  mem       REG              253,0     144776    1867778 /lib64/ld-2.5.so
rsyslogd   5380        root  mem       REG              253,0    1718232    1867780 /lib64/libc-2.5.so
rsyslogd   5380        root  mem       REG              253,0      23360    1867787 /lib64/libdl-2.5.so
rsyslogd   5380        root  mem       REG              253,0     145872    1867797 /lib64/libpthread-2.5.so
rsyslogd   5380        root  mem       REG              253,0      85544    1867815 /lib64/libz.so.1.2.3
rsyslogd   5380        root  mem       REG              253,0      53448    1867801 /lib64/librt-2.5.so
rsyslogd   5380        root  mem       REG              253,0      92816    1868016 /lib64/libresolv-2.5.so
rsyslogd   5380        root  mem       REG              253,0      20384    1867990 /lib64/rsyslog/lmnsd_ptcp.so
rsyslogd   5380        root  mem       REG              253,0      53880    1867802 /lib64/libnss_files-2.5.so
rsyslogd   5380        root  mem       REG              253,0      23736    1867800 /lib64/libnss_dns-2.5.so
rsyslogd   5380        root  mem       REG              253,0      20768    1867988 /lib64/rsyslog/lmnet.so
rsyslogd   5380        root  mem       REG              253,0      11488    1867982 /lib64/rsyslog/imfile.so
rsyslogd   5380        root  mem       REG              253,0      24040    1867983 /lib64/rsyslog/imklog.so
rsyslogd   5380        root  mem       REG              253,0      11536    1867987 /lib64/rsyslog/imuxsock.so
rsyslogd   5380        root  mem       REG              253,0      13152    1867989 /lib64/rsyslog/lmnetstrms.so
rsyslogd   5380        root  mem       REG              253,0       8400    1867992 /lib64/rsyslog/lmtcpclt.so
rsyslogd   5380        root    0r      REG                0,3          0 4026531848 /proc/kmsg
rsyslogd   5380        root    1u     IPv4         1200589517        0t0        TCP 10.10.10.90 t:40629->10.10.10.90:5514 (ESTABLISHED)
rsyslogd   5380        root    2u     IPv4         1200589527        0t0        UDP *:45801 
rsyslogd   5380        root    3w      REG              253,3   17999744    2621483 /var/log/messages
rsyslogd   5380        root    4w      REG              253,3      13383    2621484 /var/log/secure
rsyslogd   5380        root    5w      REG              253,3       7180    2621493 /var/log/maillog
rsyslogd   5380        root    6w      REG              253,3      43321    2621529 /var/log/cron
rsyslogd   5380        root    7w      REG              253,3          0    2621494 /var/log/spooler
rsyslogd   5380        root    8w      REG              253,3          0    2621495 /var/log/boot.log
rsyslogd   5380        root    9r      REG              253,3 1064271998    2621464 /var/log/custom-application.monolog.log
rsyslogd   5380        root   10u     unix 0xffff81081fad2e40        0t0 1200589511 /dev/log  

You can see that there are nowhere near 28 logfiles actually being read.

I really had to get one file monitored, so I moved it to the top, and it picked it up, but I'd like to be able to monitor all 28+ files, and not have to worry.

OS is

Centos 5.5 

Kernel 2.6.18-308.el5

rsyslogd 3.22.1, compiled with:
    FEATURE_REGEXP:             Yes
    FEATURE_LARGEFILE:          Yes
    FEATURE_NETZIP (message compression):   Yes
    GSSAPI Kerberos 5 support:      Yes
    FEATURE_DEBUG (debug build, slow code): No
    Atomic operations supported:        Yes
    Runtime Instrumentation (slow code):    No

Questions:

Why is rsyslogd only monitoring a very small subset of the files? How can I fix this so that all the files are monitored?

4
  • According to the docs, the limit should be at 100 files... So, that's not it. Nov 12, 2012 at 17:28
  • Indeed. I'm still assuming that for it to be tailing the file, it should be open in lsof? Right? Right? Nov 12, 2012 at 17:34
  • That I do not know. As soon as I figure out this problem I'm having I was going to setup a test here and see. If you run rsyslogd with strace, what's it doing with those files? Nov 12, 2012 at 18:01
  • Bollocks. I can no longer reproduce the issue. Nov 12, 2012 at 18:34

3 Answers 3

5
+100

I know some of these are obvious, but here come the things I would try...

  • Verify that the state file names are unique
  • Verify that every $InputFileName points to an existing regular file
  • Give it some time. The default polling interval is 10 seconds which could feel like an eternity
  • Clearly demonstrate what is not working. Your question says that the files are not being monitored based solely on the output of lsof. You don't mention if you actually tried writing to one of the problematic files, waited 10 seconds, and nothing showed up in the output log. And also demonstrate how you verify the files that are being monitored correctly.
  • Try to take out some of the files being monitored. Maybe it is a problem with only one of the monitored files which makes rsyslog ignore the rest of the files, too. It would help to know which one it is.

And if it helps, I have a CentOS 5.5 with the same version of rsyslog. And I tried it out with 40 files and rsyslog was able to monitor all of them. So I know that it can work correctly.

2

The question already has an accepted answer, but in case someone else needs this.

According to the rsyslog documentation the default value for PersistStateInterval is 0 ($InputFilePersistStateInterval in the context of the imfile module) meaning that the rsyslog file is only written to when the inputfile is closed. Meaning that it will appear to not work if you have a continuously open log file.

If you run into this problem, try setting $InputFilePersistStateInterval to a low value (for testing only), say 2 or 3, then once the input file has that number of lines written to it, they will be passed on to rsyslog.

0

I can't say if it is a bug or is intended to work this way. The file can be monitored even if it isn't opened. Calling stat(2) and checking mtime/size will give hints whether there is any change since the last read. At least this is how nxlog works and avoids running out of file descriptors if you need to monitor a lot more than 28.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .