Website is not loading fully sometime

[mithuuu85]

Hi,

I am facing issue with my website hosted behind F5, it works properly but sometime page is not loading fully, based on the issue we took the packet capture and can see the logs "TCP zerowindow" message is senting by F5 to Server and "TCP window full" message is sending by Server to F5 and F5 is reset the connection after this message. We did the fine-tuning for TCP profile but still the same issue, Kindly need support to solve this issue.

[Lidev]

Hi mithuuu85,

What TCP profile you use ? (tcp, tcp-lan-optimized, tcp-progressive and so on) and what tunning did you make on the TCP Profile ?

[mithuuu85]

Hi,

Initially i have used default TCP profile "TCP" , Then used TCP-wan-optimized, TCP-Lan-Optimized and F5-TCP-Mobile profile are used but none of them worked,then we configured custom TCP profile by increasing "Zero window timeout" and "Retransmit threashold" value but no luck.

[Lidev]

When you increase the Zero Window Timeout on F5 BIG-IP TCP profile, did you respect the timeout values in multiples of 5000 ?

Have you try to activate the Indefinite parameter ?

[mithuuu85]

Hi ,

We increase the Zero Window Timeout: 45000 milliseconds and how to activate Indefinite parameterm , can you please guide me on this.

Regards,

Midhun P.K

[Lidev]

On TCP Profile -> Time Management

[mithuuu85]

Hi ,

Yes, We already did this change before activate indefinite , but the issue appeared again,

[JG]

You have only shown your observations of the traffic between F5 and the back-end application servers. What were the characteristics of the traffic between your client and F5?

[mithuuu85]

HI,

Below are the Packet observed between Client to F5 and to server during the issue occurrence time.

1) Server responses with TCP Window Full

2) Then F5 respond back with TCP Zero Window from self-ip

3) we can see RST which is PSH ACK from the virtual server to the client

4) Then self-ip resets the connection to pool member due to rst from client.

Also we observed if the page was not loading fully and if we do some changes on the Virtual server after that page will open properly looks like new session/Cache clearing will make to work again.

This issue is not coming when we bypass the F5, It is working perfectly fine without any problem .

[JG]

I am afraid this description is not clear/detailed enough to show what is really happening in the traffic of the client <-> F5, and F5 <-> app server.

[mithuuu85]

HI ,

I can attach the packet capture file , Please go through that and feedback your comment on this issue.

[mithuuu85]

Hi ,

AM facing issue to upload log file in zip format, Is there any way to share the packet capture logs to you, please update

[JG]

You can get a screenshot of the Wireshark window.

[mithuuu85]

Hi,

Attached the screenshot of the Wireshark window.

[mithuuu85]

Hi,

Attached the screenshot for wireshark window

[mithuuu85]

Hi,

Below are the details about the IP address.

Client IP - 10.1.2.80

VIP IP - 10.1.221.150

Self IP - 10.1.221.85

Server IP - 10.1.221.152

[JG]

I don't see anything conclusive from the above data.

What sort of the client was involved, and did this occur to more than one client?

[mithuuu85]

Hi,

client is normal desktop and we are facing this issue with all the client accessing the VIP.

[m_soe]

You can also approach the issue from the client-side from Layer 7 perspective. Open Developer's tool bar from the web browser. Check Network traces and Console logs while going through 1) F5 virtual, 2) Direct, and compare.

Dev Toolbar:

Firefox Menu > Web Developer > Network

Chrome Menu > More Tools > Developer Tools

[mithuuu85]

Hi ,

We Check this issue from client side and got the message as below. but this setup is working fine when we bypass the F5 and access the site .

GET https://10.1.221.150/sites/all/themes/bootstrap/css/libs.min.css net::ERR_CONNECTION_RESET 200 (OK)

(index):117 GET https://10.1.221.150/sites/all/themes/bootstrap/css/rtl.css net::ERR_CONNECTION_RESET 200 (OK)

(index):468 GET https://10.1.221.150/sites/all/modules/jquery_update/replace/jquery/3.1/jquery.min.js?v=3.1.1 net::ERR_CONNECTION_RESET 200 (OK)

jquery-extend-3.4.0.js?v=3.1.1:112 Uncaught ReferenceError: jQuery is not defined

  at jquery-extend-3.4.0.js?v=3.1.1:112

(anonymous) @ jquery-extend-3.4.0.js?v=3.1.1:112

jquery.once.js?v=1.2:79 Uncaught ReferenceError: jQuery is not defined

  at jquery.once.js?v=1.2:79

(anonymous) @ jquery.once.js?v=1.2:79

drupal.js?pv505k:5 Uncaught ReferenceError: jQuery is not defined

  at drupal.js?pv505k:5

(anonymous) @ drupal.js?pv505k:5

(index):473 GET https://10.1.221.150/sites/all/themes/bootstrap/js/libs.min.js?pv505k net::ERR_CONNECTION_RESET 200 (OK)

custom.js?pv505k:154 Uncaught ReferenceError: $ is not defined

  at custom.js?pv505k:154

(anonymous) @ custom.js?pv505k:154

theme.init.js?pv505k:51 Uncaught ReferenceError: jQuery is not defined

  at theme.init.js?pv505k:51

(anonymous) @ theme.init.js?pv505k:51

(index):478 Uncaught ReferenceError: jQuery is not defined

  at (index):478

(anonymous) @ (index):478

(index):282 GET https://10.1.221.150/sites/all/themes/bootstrap/img/bg-hero111.jpg net::ERR_CONNECTION_RESET 200 (OK)

Sheet1?:embed=y&:showVizHome=no&:host_url=https%3A%2F%2Fpublic.tableau.com%2F&:embed_code_version=3&:tabs=no&:toolbar=yes&:animate_transition=yes&:display_static_image=no&:display_spinner=no&:display_overlay=yes&:display_count=yes&publish=yes&:loadOrderID=1:2 Active resource loading counts reached a per-frame limit while the tab was in background. Network requests will be delayed until a previous loading finishes, or the tab is brought to the foreground. See https://www.chromestatus.com/feature/5527160148197376 for more details

[m_soe]

Is 10.1.221.150 a virtual server ip?

[mithuuu85]

Yes 10.1.221.150 is the vip ip.

[m_soe]

Since the server IP is 10.1.221.152, it's a one-arm setup. Have you also enable "SNAT" on the virtual?

[m_soe]

Can you share the config here? Output from the following commands. Sanitize them if you need.

# tmsh ltm list virtual <virtual_name>

# tmsh ltm list pool <pool_name>

[mithuuu85]

Autonat has been enabled .

ltm virtual DATA_P_443_VS {

  destination 10.1.221.150:https

  ip-protocol tcp

  mask 255.255.255.255

  persist {

    source_addr {

      default yes

    }

  }

  pool DATA_P_443_POOL

  profiles {

    clientssl {

      context clientside

    }

    http { }

    serverssl {

      context serverside

    }

    tcp { }

  }

  rules {

    F5_Support

  }

  source 0.0.0.0/0

  source-address-translation {

    type automap

  }

  translate-address enabled

  translate-port enabled

  vlans {

    VLAN-221

  }

  vlans-enabled

  vs-index 73

}

======================================

ltm pool DATA_P_443_POOL {

  members {

    10.1.221.152:https {

      address 10.1.221.152

      session monitor-enabled

      state up

    }

  }

  monitor https_443 

}

[JG]

Are all your clients in VLAN-221?

[mithuuu85]

Hi,

Clients accessing VIP are in different vlan and we have different vlan in F5 , Few servers are hosted in Vlan-221 and few are in other vlan , But we face issue only for a specific VIP, others are working fine.

[JG]

For the sake of troubleshooting, enable access for all vlans and see if the issue persists.

[mithuuu85]

We got some findings, The same site is working fine on another F5 which is using Version 12 .

I think looks like bug issue in version 13 which is not working .

[JG]

Packets were being lost/retransmitted between the client and 10.1.221.150. Anything meaningful you can spot in /var/log/ltm ?

What's the config of you VS?

[mithuuu85]

VS setup is standard with no SSL Offload and 1 pool member.

There is no log scene on /var/log/ltm, Logs we observed on the packet capture is "TCP zerowindow" message is senting by F5 to Server and "TCP window full" message is sending by Server to F5 then keep-alive message is senting by Server to F5 but F5 is reset the connection after this message. 

[JG]

You can change the VS type to "Performance (Layer 4)" and see if that makes a difference.

[mithuuu85]

We did that as well, but same issue exist

[DerlisGarcia7]

Hi mithuuu85, i'm experimenting the same behavior in client enviroment.

How you finally resolved this issue!?