The first release-candidate of Hobbit 4.0 was uploaded to hobbitmon.sourceforge.net a few minutes ago.
This fixes several serious bugs, especially in the alert handling module where use of the DURATION specification in beta6 would delay an alert for 24 hours, and the handling of macros was very broken.
A number of platform compatibility issues have been resolved, especially on NetBSD and AIX.
The data collected for vmstat- and netstat-reports has changed, and this unfortunately means that the vmstat- and netstat-RRD files in RC1 are incompatible with previous versions, and also with files generated by LARRD. You must delete files in the old format, or collection of these data will fail. The larrd-data.log and larrd-status.log files will contain error-messages if this happens.
Some joker disabled all of the tests on my demonstration site. So I decided to move the "Enable/Disable" and "Acknowledge" CGI scripts to a secured area requiring a login. This is now part of Hobbit and can be enabled during configuration.
Several other minor bugs and improvements, the full list is below.
Thanks to everyone testing Hobbit - there are now more than 100 people on the mailing list, which is much more than I had expected. I hope you'll try out this version, so any remaining bugs can be squashed.
Regards, Henrik
Changes from beta-6 -> RC-1
NOTE: The netstat RRD file layout has changed. You must delete all "netstat.rrd" files, or your data collection will fail. Sorry, but the old LARRD format cannot cope with reality, where some systems report packet counts, some report byte counts, and some have both.
NOTE: The vmstat RRD file layout has changed, for data collected from Linux-based systems. You must delete all such "vmstat.rrd" files, or the data collection will fail. There were several incompatible formats of linux vmstat RRD files - now they all use the same format.
Hobbit should now work on AIX, and possibly other platforms that enforce strict X/Open XPG4 semantics for the ftok() routine. If you were seeing messages like "Could not generate shmem key..." or "Cannot setup status channel", then these should now be fixed. Thanks to Chris Morris for being patient with me while I struggled with the finer details of SVR4 IPC.
A new "--test" option was implemented for hobbitd_alert. Running "hobbitd_alert --test HOSTNAME SERVICENAME" will print out the rules that match the HOSTNAME/SERVICENAME combination, so you can see what alerts are triggered. (Note: man-page is not updated with this info yet).
Macros in hobbit-alerts.cfg were broken. Completely.
Specifying time-values in hobbit-alerts.cfg as "10m" was interpreted as "10 months" instead of "10 minutes". Since minutes is much more likely to be useful, the support for "months" and "years" was dropped, and "10m" now means minutes.
Specifying any kind of duration meant that the alert would not happen until 24 hours had passed (and the alert still was active).
Two new keywords in the hobbit-alerts.cfg file: STOP on a recipient means that Hobbit stops looking for more recipients after the current one has matched. UNMATCHED on a recipient means that this recipient only gets an alert if no other recipients got an alert (for setting up a default catch-all rule).
A simple syntax for "all hosts" is now "HOST=*" in hobbit-alerts.cfg
Configuring the Hobbit URL's as a root URL - "http://host/" - now works, and yields correct links to the menu, the GIF- files, and the documentation.
The Enable/Disable script (maint.pl) now picks up the menu correctly.
The "configure" script now allows you to select a second CGI directory for administration scripts (the ack-script and enable/disable script). This second directory is access-controlled; the default Apache configuration was updated to show how.
The ~/server/starthobbit.sh symbolic link was not being re-generated if the file already existed (on some platforms).
The hobbit-tips.html file is now generated during installation, so it has correct links to the icons. Previously it assumed that the icon-files had been copied to the www/help/ directory.
vmstat data for the "I/O wait cpu time" on Linux is now being collected. Linux RRD data-formats made identical across the various formats that are in use; this means any existing vmstat.rrd files from Linux systems must be deleted.
Support for NetBSD 2.0 in vmstat and netstat RRD handler.
HTTP authentication strings are now URL un-escaped, so you can use any byte-value in the username and/or password.
The info-column now includes the current address assigned to a DHCP host.
A new "noinfo" tag can be used to suppress the "info" column on selected hosts.
bbhostgrep no longer segfault's if you do not use the BBLOCATION environment setting. This kept some extension scripts from working.
If every single network test failed, bbtest-net could loop indefinitely. Seen on NetBSD.
If the number of available file-descriptors (ulimit -n) was exceeded, bbtest-net would loop forever. Seen on NetBSD.
When running http-tests via a proxy that requires authentication, the authentication info could not contain a ":" or a "@". The authentication info is now un-escaped using the normal URL decoding routine, so these characters can be entered as %XX escaped data (e.g. to put a "@" in the password, use "%40").
LDAP tests now try to select LDAPv3 first, and fallback to LDAPv2 if this is not possible. Some newer LDAP servers refuse to talk v2 protocol.
Memory debugging was accidentally enabled in beta6 - it is now disabled.