PfSense wakes my Synology NAS boxes from hibernation every 1 hour….

pfcode

No. Didn't work, re-created it, but still wake up after 1 hour, but I only created it at pfSense side. would try to enter the manual IP, rather than auto DHCP, at the NAS side

Derelict

Me? I'd put the NAS on a mirror port on my switch, let it sleep, then run a packet capture until it's woken up and see what caused it.

Why guess at the problem?

johnpoz

Agreed take a sniff and see exactly what is causing a wake up.

pfcode

@johnpoz:

Agreed take a sniff and see exactly what is causing a wake up.

As my NAS hibernation logs tells, also from Synology tech support:

It appears that the router keeps trying to distribute DHCP addresses to the units, even though they already have an IP address.

I'm not sure its the NAS problem or pfSense problem, NAS is sending DHCPREQUEST command to pfSense every 1 hour, but if the lease time isn't expired, Shouldn't pfSense NOT distribute DHCP address to the unit?

doktornotor

@pfcode:

It appears that the router keeps trying to distribute DHCP addresses to the units, even though they already have an IP address.

Yes, of course, the DHCP server keeps "distributing" IP address requested by your NAS, kindly cf.


dhcpd: DHCPREQUEST <my nas="" lan="" ip="">from <my nas="" mac="" addr="">via em1</my></my>

Also, the DHCP leases expire and need to get renewed (after a configurable period. check the GUI). So yeah, this will happen "even though they already have an IP address", perhaps the Synology support could educate themselves a bit.

Finally, if you do not want DHCP, then configure a static IP on your NAS and move on!

hda

@pfcode:

… but I only created it at pfSense side ...

Before you jump in mirroring investigation, if you know-how. I am not sure if you are exact & correct. You really should test giving the NAS a new number manually (outside the range/pool) AND manually input it in pfSense too.
This is to assure yourself the volatile memory about .107 is flushed on both sides… Maybe even reboot both hardware.

Derelict

Yes. It should not be a DHCP static mapping. That still has to periodically renew. It should be a static IP address directly configured on the NAS, outside the DHCP pool.

johnpoz

What dhcp server doesn't just out of the blue send out hey you want to renew your lease packets.. But if it gets a request then yeah it will send back something..

Tell their their Tech Support to tell you how not to send a request.. What is the lease time to start with 2 hours? Normally a dhcp client will request renewal of its lease at the 50% mark. So if you want the nas not to request, then it should be long lease. If the device is in standby/sleep why would it send out a dhcprequest is the question I would have.

Derelict

True. Should stop guessing and sniff.

doktornotor

@johnpoz:

If the device is in standby/sleep why would it send out a dhcprequest is the question I would have.

This is not really any suspend-to-disk/ram hibernation. The only thing "hibernating" are the HDDs which are spun down. Now, look at this:


# ps ax | grep dhclient
 4606 ?        Ss     0:00 /usr/sbin/dhclient -4 -d -q -lf /etc/dhclient/ipv4/dhcpv4.leases.eth0 -pf /etc/dhclient/ipv4/dhcpcd-eth0.pid eth0
 5551 ?        Ss     0:00 /usr/sbin/dhclient -6 -pf /tmp/dhclient6-eth0.pid -lf /etc/dhclient/ipv6/dhcpv6.lease.eth0 -cf /etc/dhclient/ipv6/dhcpv6.conf.eth0 -nw eth0

Those dhcpv[46].lease.eth? files have stuff like this:


lease {
  interface "eth0";
  fixed-address 192.168.1.10;
  option subnet-mask 255.255.255.0;
  option routers 192.168.1.1;
  option dhcp-lease-time 7200;
  option dhcp-message-type 5;
  option domain-name-servers 192.168.1.1;
  option dhcp-server-identifier 192.168.1.1;
  option host-name "DiskStation";
  option domain-name "example.com";
  renew 2 2015/05/19 19:46:58;
  rebind 2 2015/05/19 20:35:47;
  expire 2 2015/05/19 20:50:47;
}

This is just stupid. There's /tmp which is on tmpfs (ramdisk). Now, the lease file is on /etc instead, which is a HDD. Of course, when the lease is renewed, the HDD needs to spin up to write to the file.

Tell Synology to move their temporary dhcpv[46].lease.eth? shit to /tmp when people enable "hibernation" – and it won't wake up HDDs from hibernation. ::)

johnpoz

the hdd are just spun down - ah yeah wouldn't call that sleep anything ;)

If that lease file is from actual nas then looks to be only 2 hour lease.. That is not the default lease time in pfsense dhcp server is it? Thought it was like 24, yup quick look on dhcp server tab "The default is 86400 seconds." So the nas is only requesting 2 hours by default?

doktornotor

@johnpoz:

If that lease file is from actual nas then looks to be only 2 hour lease.. That is not the default lease time in pfsense dhcp server is it? Thought it was like 24, yup quick look on dhcp server tab "The default is 86400 seconds." So the nas is only requesting 2 hours by default?

The default lease time is 7200 seconds on pfSense; the default maximum lease time is 86400 secs (one day).

johnpoz

you are correct as always dok

Lease Obtained. . . . . . . . . . : Tuesday, May 19, 2015 3:23:15 PM
Lease Expires . . . . . . . . . . : Tuesday, May 19, 2015 5:23:15 PM

doh – helps if you actually read what your doing a copy of. quick enough fix ;)

Lease Obtained. . . . . . . . . . : Tuesday, May 19, 2015 3:23:15 PM
Lease Expires . . . . . . . . . . : Wednesday, May 20, 2015 3:28:28 PM

Now its 24 hours..

Derelict

I might have read something somewhere about clients requesting renewal halfway through their lease.

pfcode

Thanks All, by assigning a manual IP in the NAS network configuration, solved the issue, It was 'get network configuration automatically (DHCP)' before. I will let the Synology tech guys to investigate further to why it is sending dhcprequest every 1 hour (3600 seconds) to the router.

doktornotor

@pfcode:

I will let the Synology tech guys to investigate further to why it is sending dhcprequest every 1 hour (3600 seconds) to the router.

Well, because you did not configure any default lease time, so you get 2 hours by default, since Synology DSM does not supply any lease time in the DHCP request… That's not what wakes it up, there'd be no problem with that if they were not writing the lease file to the HDD instead of tmpfs.

NOYB

@pfcode:

@johnpoz:

Agreed take a sniff and see exactly what is causing a wake up.

As my NAS hibernation logs tells, also from Synology tech support:

It appears that the router keeps trying to distribute DHCP addresses to the units, even though they already have an IP address.

I'm not sure its the NAS problem or pfSense problem, NAS is sending DHCPREQUEST command to pfSense every 1 hour, but if the lease time isn't expired, Shouldn't pfSense NOT distribute DHCP address to the unit?

Dude or Dudette. Whichever the case may be. Why do you continue asking question with you've been given the answer?

The definitive answer to your issue is in the sniff. No one can tell you (without guessing) what is waking your NAS without visibility to what is on the wire. Sniff it and report back with the capture.

If you want to discuss DHCP protocol, I suggest the DHCP and DNS forum.

pfcode

@doktornotor:

@johnpoz:

If the device is in standby/sleep why would it send out a dhcprequest is the question I would have.

This is not really any suspend-to-disk/ram hibernation. The only thing "hibernating" are the HDDs which are spun down. Now, look at this:
# ps ax | grep dhclient
 4606 ?        Ss     0:00 /usr/sbin/dhclient -4 -d -q -lf /etc/dhclient/ipv4/dhcpv4.leases.eth0 -pf /etc/dhclient/ipv4/dhcpcd-eth0.pid eth0
 5551 ?        Ss     0:00 /usr/sbin/dhclient -6 -pf /tmp/dhclient6-eth0.pid -lf /etc/dhclient/ipv6/dhcpv6.lease.eth0 -cf /etc/dhclient/ipv6/dhcpv6.conf.eth0 -nw eth0
Those dhcpv[46].lease.eth? files have stuff like this:
lease {
  interface "eth0";
  fixed-address 192.168.1.10;
  option subnet-mask 255.255.255.0;
  option routers 192.168.1.1;
  option dhcp-lease-time 7200;
  option dhcp-message-type 5;
  option domain-name-servers 192.168.1.1;
  option dhcp-server-identifier 192.168.1.1;
  option host-name "DiskStation";
  option domain-name "example.com";
  renew 2 2015/05/19 19:46:58;
  rebind 2 2015/05/19 20:35:47;
  expire 2 2015/05/19 20:50:47;
}
This is just stupid. There's /tmp which is on tmpfs (ramdisk). Now, the lease file is on /etc instead, which is a HDD. Of course, when the lease is renewed, the HDD needs to spin up to write to the file.

Tell Synology to move their temporary dhcpv[46].lease.eth? shit to /tmp when people enable "hibernation" – and it won't wake up HDDs from hibernation. ::)

This is a good catch. Thanked.