Hi there, I'm having some trouble on the reports that Ganglia generates, it's reporting inaccurate data from the actual system data. I'm running FreeBSD 5.4 and Ganglia is compiled from the ports tree (version 3.0.1). I'd really appreciate any help you can give.
The web frontend tells me: Hardware CPUs: 4 x 62.92 Ghz Memory (RAM): 4094.99 GB Local Disk: Unknown Most Full Disk Partition: 100.0% used. Software OS: FreeBSD 5.4-RELEASE-p4 (i386) Booted: July 25, 2005, 3:22 pm Uptime: 28 days, 1:48:48 Swap: Using 0.0 of 4096.0 MB swap. But in fact I have 4GB of RAM and 2 x 3.2Ghz CPU: # sysctl hw.physmem hw.physmem: 3212009472 (Note the physical ram is actually 4GB, but Freebsd only utilises 3GB at the moment) # dmesg | grep -i cpu CPU: Intel(R) Xeon(TM) CPU 3.20GHz (3191.95-MHz 686-class CPU) Hyperthreading: 2 logical CPUs FreeBSD/SMP: Multiprocessor System Detected: 4 CPUs Also, the second thing Ganglia reports incorrectly is the disk usage. On the web frontend it tells me that disk_total is 30GB and disk_free is 0GB, whereas in reality: # df -a Filesystem 1K-blocks Used Avail Capacity Mounted on /dev/amrd0s1a 2026030 920356 943592 49% / devfs 1 1 0 100% /dev /dev/amrd0s1d 1012974 32 931906 0% /tmp /dev/amrd0s1f 24238446 21218026 1081346 95% /usr /dev/amrd0s1e 3045006 1801016 1000390 64% /var However when NFS mounts are used, it reports something completely differently, taking fully into account of the NFS disk free spaces etc., that the disk_total is reported as 49.64GB and disk_free as 2.22GB; in reality, it is: $ df -h Filesystem Size Used Avail Capacity Mounted on /dev/amrd0s1a 1.9G 390M 1.4G 21% / devfs 1.0K 1.0K 0B 100% /dev /dev/amrd0s1d 989M 62K 910M 0% /tmp /dev/amrd0s1f 23G 18G 3.0G 86% /usr /dev/amrd0s1e 2.9G 906M 1.8G 33% /var kettle:/usr/local/content 23G 20G 1.0G 95% /usr/local/content kettle:/usr/ports/packages 23G 20G 1.0G 95% /usr/ports/masterpackages My guess is that when find_disk_space() is called, disk_free is calculated from most_full partition, which in this case is devfs. However devfs of FreeBSD is like /proc of Linux and reporting its size is pretty much meaningless in this context. But I really don't understand why disk_free gives 0 normally and 2G when NFS is mounted. I'm running Ganglia across a 5 freebsd webserver cluster and all of them are reporting the same set of data. Could someone please help me to get Ganglia to report the correct values? Many thanks! Kind regards, Hui This email may contain confidential material. If you were not an intended recipient, please notify the sender and delete all copies. We may monitor email to and from our network.
