Cisco Application :: CSS 11503 Flow Idle Timeout Not Working As Expected?
Jan 20, 2012
I have a CSS 11503 with a basic content rule for TCP 10000 going to a few backend servers. I was looking into the default timeout values for flows and when testing using telnet the flow didn't terminate as expected?
For example, i have no 'timeout multiplier' specified in the config and when i look at the output of 'show flow-timeout default' it tells me the default 16 seconds timeout is in effect for *. With that in mind, i telnet to the content rule vip on TCP 10000 and on the backend server using wireshark i can see the TCP threeway handshake. With no data passing i'd expect the CSS to terminate this flow after 16 seconds.. yet it takes exactly 128 seconds before wireshark shows the RST and the flow is terminated. 128 being 8 times the default 16 second flow timeout.
If i try to force the connection to close early by specifiying 'flow-timeout-multiplier 2' in the content rule, or even a multiplier of 40, it still waits 128 seconds to close the telnet connection.
We have a pair of CSS 11503 installed in our DC. Stickiness is configured for one of the application since long back and was working pretty fine till last couple of months. Since last two months, we observed that CSS is not distributing sessions the way it suppose to be. Mostly, it forwards the session to same server even though request is coming from different sources. Once we refresh the sessions manually, it starts working fine. We have to do this exercise manually every alternate day.
I am looking for the way to define an idle timeout for specific flows on an ASA5580 by using Cisco security manager. For ex I needed to define a specific idle timeout for connections beetween specific devices (Devices in vlan1, Device2 in vlan2).To test it I did following changes by CLI and it works fine. access-list L1 extended permit ip <@IP1> <mask1> host <@IP2> class-map CM1 match access-list L1 policy-map PM1 class CM1 set connection timeout idle 02:00:00
I try do do the same configuration with CSM in order to be able to manage each changes only by using CSM.So I defined Access control list, Traffic flow and then I define timeout in CSM --> PIX/ASA/FWSM Platform --> Service Policy Rules --> IPS, QoS and Connections Rules -> connections settings -> Traffic flow idle time-out. The problem is that each time I deploy the configuration with CSM I loose the timeout config line which is the most important for my application..
On the modem page with all the settings Under basic settings For connection it says Always Connected The box underneath says idle timeout 5 (that's in minutes)
I am unable to change that number or even get a cursor to appear in that box (I want to change it to a 0 - my internet connection has been dropping out when going idle). I can't right click or anything. That is using Firefox. When I say dropping out, I mean, the 3 computers on the network become unuseable - nothing works and things don't appear to be connected to the internet even though my green ADSL light stays on like it's connected - it is not.
I tried the same modem page in IE, and that box with the 5 in it is still showing 5, but this time it's just grayed out - again, can't be changed. The firmware updates are all upto date (apparently - according to it's check)
I have ASA 5510 with 8.2.4 and 8.0.x OS and all seem to have common problem of idle TCP connections not timing out. The host to host connections are coming over VPN tunnels. I have default timeouts on all the firewalls. I have tried changing global timeouts and as well as host specific timeouts using MPF but doesn't work at all ! The problem is when TCP connections are sitting idle in conn table for days and when connection limit of 50,000 conns reach the firewall starts behaving unpredictably dropping packets or unresponsive! I need the unused idle connections to timeout which is NOT happening either by changing global values or MPF.
If i set up a pptp vpn between a Cisco rv082 router and a microsoft client,Can i set the client idle timeout someway? or Have a default value pre- configured for this?Because this device support 5 users to connect at the same time. It would be best for me, if the device drop the client if it does not use the tunel.
I have an SG300-20 here for testing (firmware: 22.214.171.124, boot version: 126.96.36.199, language version: 188.8.131.52 English). Everything seems to work on it, except, that if I choose Radius authentication by mac address only, then the switch does not honor the Idle-Timeout and Session-Timeout attributes from the Radius server (freeradius).
The setup is the following: I have a no name access point plugged in to switch port gi1. The port gi1 is set up for Radius authentication by mac address only. The access point itself is authenticated, no problem with that. If I connect through the access point by (say) a mobile phone, it is authenticated, no problem. The radius server does send the Idle-Timeout and Session-Timeout attributes, I checked it by running "freeradius -X", both are set to 30 seconds. Then I turn off the wireless card in my mobile phone and check the dot1x users by "show dot1x users". My mobile phone's mac address remains there for 5-10 minutes, so the Idle-Timeout and Session-Timeout does not work.
Another way I could resolv this problem is by explicitely asking the switch to reauthenticate the user. Unfortunately there is no CLI command to do just that, I can do however a reauthentication on a port using "dot1x re-authenticate gi1" (for example). But it does not work as it is expected: the switch uses the stored mac-address to reauthenticate the user, so nothing changes on the port (unless something changes in the radius server). I think it should work like the following: remove the authenticated user from the port, and whenever that mac address makes some network traffic, then reauthenticate as if it were a completely new connection. BTW: it would work for me also if I could just remove an authenticated user from a port, but I did not find a command to do that.
As a last resort I can simply shutdown the port, bring it up again ("shutdown" and "no shutdown" in the interface config), then all users are removed from the port and they all mush reauthenticate. But it causes a network outage for a couple of seconds for all users on that port, on a busy access point it is quite disturbing, and it is not an elegant way to do this.
So my actual question is: is there a way to remove an authenticated user either automatically (Idle-Timeout and Session-Timeout) or manually from this switch?
I enclose the relevant part of the running config.
On my 1841 when i enter the "ip flow-cache timeout active 2" command it accepts this command with no errors. But when i look at my running-config this does not list.I did the same thing on my 2811's and 3745 and it shows up in the running-config. Should I assume if it doesnt' show up in my config file than it is not applied? How can I verify that it is or isn't?
I have a number of web sites that are currently being load balanced by CSS 11503s runninng 8x code. I was recently requested to configure HTTP --> HTTPS redirects on the CSS for every site. In the past, I have only configured the redirects for sites that had a requirement. Now it appears that the server teams want all content encrypted.
1) What impact will this have on the CPU? 2) What impact will this have on Memory utilization? 3) Is there a maximum nubmer on redirects? 4) Are there other things I should be concerned about?
i have two CSS-11503 in redundant mode running 8.20 code. We had an incident in our network where a layer 2 loop caused some high traffic through the CSS' and had to shutdown some network gear(including the CSS) to clear the problem. When the CSS' were powered back up, the SSL service was suspended, why this would occur? There rest of the config appeared normal. I am the only person on these boxes, the configs were written, and I have never had a reason to suspend the ssl service.
We now have a new requirement . We are replacing existing pair of CSS with ACE 4710 appliances. The problem here is that I can see from the configuration that some SSL certificate installed in CSS .Is it possible to transfer the existing SSL certificate from the 11503 to the ACE? Or, do we need to generate a new key pair and CSR on the ACE? Is there any document available to know the steps for the same.
We have several pairs of CSS11501 and 11503 in our network.This issue affects only one pair of CSS11503 in one of our data centres. [code] We use vrrp in one-armed mode for load balancing and they units have performed great for a number of years. We're obviously going to be migrating to ACE ... but not just yet.We have started to experience a problem with replicating the configurations between two CSS11503 in a pair.When running the commit-VipRedundConfig, it starts off happily enough, though slowly.Ending with "working" and the spinning cursor, even after 1 hour the script hasn't completed.We noted on the backup CSS that the APP configuration disappears during the process and I can't remember if this is normal behaviour.
Re-adding the app session configuration seems to interrupt the process, and when checking the configuration on the backup CSS approximately half of it is missing. Everything after the first owner is gone.
1. Configuration is too large, or just large enough to make the commit script take too long for realistic service. 2. Software bug? 3. Combination of both. 4. From now on manually add config to both CSS's and maintain it by process management.
I only have configured load balancing on apache with a very simple setup. I have to deploy 2 applications on my clients environment that run inside jboss. One of these applications needs session to be sticky to work properly. The other does not.
In apache I can configure is the sticky parameter is true or false, based on the url, like /appA/* is sticky and /appB/* is not sticky. Can I do that in a CSS 11503? My client insists that it is impossible. That the CSS is only ip based.
I copied the configuration below from the manual: owner arrowpoint # content ruleWapSticky
I have a single cisco 11503 load balancer.There is a single Banner student information system which is load balanced on it with Virtual ip 10.3.20.101 which is working fine without any issues .I am now trying to add an Oracle ERP application with virtual IP 10.3.20.230 and physical ips 10.3.19.22 and 10.3.19.23 all on port 8003.When I just make the group ERP-Apps-Grp active , the vitual ip address 10.3.20.230 is pingable , but when I make the the content Erp_IAT active it stops pinging. [code]
content My_Group add service blade1 add service blade2 add service blade3 vip address 184.108.40.206 advanced-balance arrowpoint-cookie
So I have 3 blades which are proxy servers and user go first to an MS ISA server then the VIP of the CSS and then the rules processes them give them a blade and chuck them out onto the Internet.
I want to leave the above rule, but remove one blade create an additional content group with that blade and have it process requests for a particular site so, I would create the following
content My_Group2 add service blade3 vip address 220.127.116.11 advanced-balance arrowpoint-cookie
So my question is can I do that having the same VIP's etc so if a request comes in and it matches www.thewebsite.com that the second content rule matches it 'better' and therefore processes it or would it still be caught by the "/*" content group. I don't want to create more VIPS as I have a real ache getting firewall rules done.
I'm not sure if my terminology is correct when using hairpinning but i was wondering if there is any special config needed when you try to access a content rule VIP from a server that's configured as a member of a source group on the same CSS?
So say i have a content rule with a VIP 18.104.22.168 and i also have two servers 192.168.1.1 and 192.168.1.2 that are part of a source group with VIP of 22.214.171.124. My problem at the moment is if from the servers 192.168.1.x i try to ping the other VIP 126.96.36.199 that's configured on the same CSS then it doesn't work and ping fails. The same happens with HTTP traffic to the 188.8.131.52 VIP.
I would have thought that the NAT of the source group would happen before the routing so the 192.168.1.x IP's would be natted to 184.108.40.206 and then passed over for routing where the CSS would see that the VIP 220.127.116.11 is local and it would send it on it's way.
I thought it might be ACL related but i increased the verbosity of acl logging and couldn't see anything in the logs.The source group works fine on it's own and from the CSS itself i can ping the 18.104.22.168 VIP fine. It just seems that from the source group members i can't ping the VIP.
access-list <#> permit/deny <protocol> <sourceAddress> <sourceMask> <destinationAdd> <destinationMask>Say I applied an ACL inbound on Fa0/0, would the source address be the outside the LAN?So if took the same ACL and applied it as outbound, would the source need to be change to an IP inside the LAN?I am a bit confused by the data flow I'm seeing in packet tracer simulation mode to. I set up an ACL for testing purposes "access-list 199 permit ip 22.214.171.124 0.0.0.63 any" set as inbound, the idea being it permits any traffic from the .0 subnet.When I watch the packet in the simulation, it makes it to the destination address then is dropped by the router on it's way back out to the sender.
I have already raised this discussion on "LAN, Switching and Routing" group. But I guess this is the right group for my queries. So I am sending my queries in this group again.
We are using CSS 11503 with one 16FE line card. We have connected 3 servers with redundant link. So FE1-2 in Server1, FE 3-4 in Server2 and FE5-6 in Server3. Our system team has configured APA in their servers as they are using HP-Ux.
1) Do we need to do any configuration at line card.
2) Do we need to do ether-channel at loadbalancer end. if yes, can you share me any cisco doc on how to do it.
I have 2 pair of 11501 switches and 1 pair of 11503 switches on 3 sites(LA, China, Taiwan).Each site has a pair of 1105x switch running as redundancy between them and is a standalone which will not interact with others.Recently a series of interfaces(ports) down happened to every active 1150x switches without any reason and log.Especially today, it happened to active switches at 5:39 AM meanwhile on 3 sites.
I have 10 new AIR-CAP3502I-A-K9 connected to a WS-C3750X-48PF-L switch. 8 of the APs power on and connect perfectly, but two are problematic. Both devices are granted power, but they never go past the stage of getting power from the switch. A look at POE shows: [code]
Where the AP that is not working is connected to Gi1/0/4. The interface shows down/down. I've tried shutting the interface, removing power, cdp etc.
Are there any other tricks you can think of to get it working or would you say that it's 2 faulty APs? I am not based at site unfortunately so I cannot console to the APs and check them out.
I'm trying to design a CSS configuration that allows servers in the same vlan to be the source and destination of load-balanced traffic. My thought is to add two new vlans, one for the VIPs and one for the servers, then NAT the source IPs going from the LB to the servers.
Is this the right way to do it?I've never NATted using CSSs, so I wanted to verify what I'm thinking.Our current config trunks the vlans -