Pfsense hangs every day because of buffers
-
Hello,
We have two pfsense 2.2.2 in HA and the active one hangs every few hours and it seems it's because of buffers, but we are not sure if it's the reason. The secondary doesn't get the active role because the active still responds to ping sometimes, we can access to it with SSH but we can't ping to internet, but for few minutes the OpenVPN tunnels that we have still responds. Other times it's freezed and we have to reset it directly. We have decided to shut down the secondary server to see if it's part of the problem, but it still happens.
They're virtualized in two separated proxmox 3.4 server as KVM with 3Gb of RAM, with 4 cores and 64Gb of disk in VIRTIO over SSD disk (in real SSD is a LVM partition). The KVM virtualization flag is on.
Right now top shows:
last pid: 62068; load averages: 0.36, 0.26, 0.19 up 0+01:54:22 08:47:12
96 processes: 2 running, 94 sleeping
CPU: 0.0% user, 0.0% nice, 0.5% system, 0.3% interrupt, 99.2% idle
Mem: 51M Active, 214M Inact, 174M Wired, 363M Buf, 2519M Free
Swap: 8192M Total, 8192M FreeWhen we see the top, the proc
I have attached two graphs. The one "pfsense-observium.png" is the graph taken from our Observium that checks the pfsense via SNMP since the last sunday and you can see how the buffers goes over 450Mb more or less and every time dissapears it's when we have rebooted it. The graph pfsense.png is the same from the pfsense RRD graphs, but doesn't show very well what happens. As you can see, we have never reached more than 15% of RAM, and it seems that the buffers began to get RAM and don't liberate it.
The processes that gets more RAM are php and php-fpm with 228M. We have 8 LANs, a captive portal, 4 openvpn servers and 5 openvpn clients, few CARP virtual IPs (but as I said before, the slave server is down).
Any suggestion?
Thanks!
-
Few minutes after starting the post, right now the top shows:
last pid: 1893; load averages: 0.06, 0.13, 0.15 up 0+02:30:55 09:23:45
96 processes: 1 running, 95 sleeping
CPU: 0.0% user, 0.0% nice, 0.3% system, 2.2% interrupt, 97.5% idle
Mem: 40M Active, 228M Inact, 175M Wired, 94M Buf, 2515M Free
Swap: 8192M Total, 8192M FreeAs you can see, in my previous post there where 363M of Buf, and now is 94M. Two minutes ago it was in 423M more or less, and we haven't done anything, it has release the buffers :o
-
What packages are you running?
-
The processes that gets more RAM are php and php-fpm with 228M. We have 8 LANs, a captive portal, 4 openvpn servers and 5 openvpn clients, few CARP virtual IPs (but as I said before, the slave server is down).
And, what exactly do you intend to do with the free RAM? Take a picture with it? "2519M Free" - what's your problem with RAM being used? Unused RAM -> wasted money.
As you can see, in my previous post there where 363M of Buf, and now is 94M. Two minutes ago it was in 423M more or less, and we haven't done anything, it has release the buffers :o
BTW, those buffers you are trying to get rid off, that's disk cache - see top(1)…
Some reading: https://www.freebsd.org/doc/en/books/arch-handbook/vm.html
-
The packages that we have installed:
-
arping
-
iftop
-
iperf
-
mtr-nox11
-
nmap
-
NRPE v2
-
OpenVPN Client Export Utility
-
pfBlockerNG
-
Quagga OSPF
We thought that "buf" means for "memory buffers" but not for disk. So, it could be a problem when trying to write into disk, couldn't it?
Any other suggestion would be appreciated. Thanks
-
-
There's no problem, let alone when writing to disk… Reading from RAM is faster than reading from disk, what's currently used gets cached. Focus your debugging efforts somewhere else.