Forum Discussion
Communications problems between servers
Hi,
We have some problems with the comunications of some server. The situation is below:
- Server A on VLAN 1 with IP 192.168.1.1/24
- Server B on VLAN 1 with IP 192.168.1.3/24
- Server C on VLAN 1 with IP 192.168.1.5/24
- Virtual Server on VLAN 2 with IP 192.168.3.2/24 (its nodes are servers A and B)
The service is over http and https, and we are not using SNAT.
When clients from Internet try to connect to the Service there is no problem; but when server C ask for virtual server service, the connections never reach the server A or B. We just have this problem with connections over servers are in the same VLAN.
If we do a tcpdump we can see:
12:00:15.139092 IP 192.162.1.5.dyna-lm > 192.162.3.2.http: S 1702227370:1702227370(0) win 65535 <mss 1460,nop,nop,sackOK>
But the connection seems it doesn´t go out from the BIG-IP since in the servers A and B we cannot see any connection from server C or Virtual Server.
Any idea? Could you help me?
Thanks Regards
33 Replies
some questions: - what is the default gateway of the servers? does that device know the route to 192.162.3.0/24? - is virtual server 192.162.3.2 enabled on vlan1? - where did you make that tcpdump and with which arguments?
without more background i could imagine the reply from server A or B won't go via the BIG-IP as it is in the same network as server C and SNAT isn't used. so server A or B might go directly to server C. have you checked that is possible and happening?
- Edu_50128
Nimbostratus
The default gateway of the servers is the BIG-IP.
Yes, this IP is enabled on both VLANs.
The tcpdump we have done is on the BIG-IP with this options:
tcpdump -i any src host 192.168.1.1 or dst host 192.168.1.1 or src host 192.168.1.2 or dst host 192.168.1.2 or src host 192.168.1.5
Thanks !!!
- Hamish
Cirrocumulus
Can you add the -e flag to tcpdump to verify which interface and the correct tagging of the packets.
H
could you make it simpler and add Hamish flag to be sure we don't miss something:
tcpdump -i 0.0 -nn -e host 192.168.1.1 or host 192.168.1.3 or host 192.168.1.5
and try again
also is it possible to capture packets on the actual servers to see if they communicate between themselves?
- Edu_50128
Nimbostratus
Now we have this on the tcpdump:
08:51:34.775753 00:1f:29:0b:21:a6 (oui Unknown) > 00:01:d7:92:e4:87 (oui Unknown), ethertype 802.1Q (0x8100), length 66: vlan 32, p 0, ethertype IPv4, 192.168.1.5.drmsfsd > 192.168.3.2.http: S 867528809:867528809(0) win 65535
In the servers (A and B), we don´t see any packet from the BIGIP or from server C, there is no connection between them.
- Edu_50128
Nimbostratus
sorry, when it indicates vlan 32 it meens vlan 1
- Edu_50128
Nimbostratus
in the tcpdump it also appears these lines:
08:51:37.741338 00:1f:29:0b:21:a6 (oui Unknown) > Broadcast, ethertype 802.1Q (0x8100), length 64: vlan 1, p 0, ethertype ARP, arp who-has 192.168.1.10 tell 192.168.1.5
08:51:37.741357 00:1f:29:0b:21:a6 (oui Unknown) > 00:01:d7:92:e4:87 (oui Unknown), ethertype 802.1Q (0x8100), length 66: vlan 1, p 0, ethertype IPv4, 192.168.1.5.drmsfsd > 192.168.3.2.http: S 867528809:867528809(0) win 65535
192.168.1.10 is the gateway for the servers, and it is the BIG-IP.
you are probably trying to hide some things from us or make it more simple, you could edit your earlier post instead though :)
you also didn't execute the command as requested i still see ports being translated in names, and possibly you still use -i any, something im not sure might have odd results, could you do it as requested?
are you using route domains? or config partitions?
also is it possible to do it without filters? is the box loaded, are there more virtual servers?
- Edu_50128
Nimbostratus
the BIG-IP we are using has thousands of connections, I´m trying to put just the connections with this machines, since we have several connections from clients (most of them)
I´m going to send you the output of -i, let me do it please
we are not using router domains or config partitions
Yes, In this BIG-IP there more Virtual Servers, more than 12...
- Beinhard_8950
Nimbostratus
Hi,
Until you have fixed a correct tcpdump that I belive will show that the request from the server c will be LB to A or B it will not work. You need to put a snat rule for connection to the vs for traffic from the server c. That because you have a asymetric situation here. when the connection will be loadbalanced to server a or b they will see that the src is from server c and on the same vlan so they will send the traffic directly to the server and bypass the f5. easy way if you have problem with the f5 tcpdump and filtering is to do a tcpdump on server c and look for the server a or b address and you should see syn-ack:s =)
Br
Beinhard
- yeah that point i also made quite early in the question. build a similar situation just now and you get some thing like this if i translate it to your IPs 192.168.1.5 --> 192.168.3.2 client side 192.168.1.5 --> 192.168.1.1 (server A for example) server side then 192.168.1.1 is going to directly contact 192.168.1.5 which fails because 192.168.1.5 isn't expecting that traffic from 192.168.1.1.
Recent Discussions
Related Content
* Getting Started on DevCentral
* Community Guidelines
* Community Terms of Use / EULA
* Community Ranking Explained
* Community Resources
* Contact the DevCentral Team
* Update MFA on account.f5.com