The Neverfail Channel allows replicated data and status/control information to pass between the servers in a Neverfail group. When the Neverfail Channel disconnects unexpectedly, it is referred to as a channel drop. Channel drops interrupt the replication of data and transmission of status/control information and can result in false failovers. This behaviour is not by design and can be caused by the issues described below.
Here are the main causes, their symptoms and the solutions that can be applied:
Symptoms
The message "java.io.IOException: An existing connection was forcibly closed by the remote host" appears in the active server's NFLog.txt file, and the channel connection between the servers is lost.
Cause
This condition is unusual and generally points to an application, or Windows itself, experiencing a fault on the passive server. The most likely issue here is a sudden reboot / restart of the passive server and may be due to one of the following causes:
Resolution
2. Passive server does not meet minimum hardware requirements
Symptoms
The data rate between the servers is very high during a Full System Check and the channel drops.
Cause
The passive server does not meet the recommended hardware requirements for Neverfail Engine or it meets them but is much less powerful than the active server. The underpowered server cannot apply the received replication data from the active server at the rate that the data is sent to the passive server.
Resolution
In order to avoid reinstalling your Neverfail Engine solution, it is best to tackle this issue by upgrading the hardware (for example, memory, CPU) on the passive server. It is important to establish the identity (Primary, Secondary, or Tertiary) of the affected server before you perform the upgrade. An upgraded Primary server may require a new Neverfail license if the HBSIG is changed. Upgrading the Secondary server will not require a new license.
3. Hardware or driver issues on channel NICs
Symptoms
The Neverfail Channel drops or disconnects and reconnects intermittently.
Cause
Resolution
5. Firewall connection
In both a LAN or WAN deployment of Neverfail Engine, the channel may be connected via one or more Internet firewalls. Since firewalls are intended to block unauthorised network traffic, it is important to ensure that any firewalls along the route of the channel are configured to allow channel traffic.
Symptoms
The Neverfail Channel cannot connect or connects and disconnects continuously.
Cause
In a WAN deployment, port #57348 (or any other port configured for the Neverfail Channel) is closed on one or more firewalls on the route between the channel NIC on the Primary server and its counterpart on the Secondary server.
Resolution
Open port #57348 (and any other port configured for the Neverfail Channel) on all firewalls on the route between the channel NIC on the Primary server and its counterpart on the Secondary server.
6. Incorrect Neverfail Channel configuration
Symptoms
Cause
Identical IP addresses at each end of the channel, IP addresses in different subnets without static routing at each end of the channel, or a channel NIC configured for DHCP when a DHCP server is not available.
During the installation of Neverfail Engine, some configuration data from the Primary server is copied to the Secondary server. This includes configuration information for any NICs. The Help text displayed on the Neverfail Engine Deployment wizard describes how to configure the IP address for each NIC on the Secondary server. If this step is not completed, it is possible for one or more channel NICs on the Secondary server to contain a variety of incorrect addresses derived from the Primary server.
For example, assume you want the following correct configuration after deployment:
Primary:
Public NIC: 192.169.1.101
Channel NIC #1: 9.0.0.1
Channel NIC #2: 10.0.0.1
Secondary:
Public NIC: 192.169.1.101
Channel NIC #1: 9.0.0.2
Channel NIC #2: 10.0.0.2
Immediately following the restore/Plug and Play phase of the Secondary installation, Channel NIC #1 on the Secondary server may have acquired any of the following IP addresses:
(a) 192.169.1.101
(b) 9.0.0.1
(c) 10.0.0.1
(d) No static IP address (i.e. NIC is configured to use DHCP)
Clearly, none of these will allow a connection to address 9.0.0.1 on the Primary server - (a) and (c) are in a different subnet, (b) is a duplicate IP address, and (d) will fail because there is normally no DHCP server connected to the channel NICs.
Which address is assigned to Channel NIC #1 on the Secondary server depends on the exact driver configuration of the NICs on that server, as compared with the NICs on the Primary server. The most likely result, and the one usually expected during deployment, would be for the IP address of Channel NIC #1 on the Primary server to be transferred to Channel NIC #1 on the Secondary server.
On rare occasions, if the Primary and Secondary servers have NICs of the same type in a different order, both the name and IP address of a channel NIC on the Primary server may be transferred to the principal (public) NIC on the Secondary; or the name and IP address of the principal (public) NIC may be transferred to a channel NIC. Similarly, the names of the channel NICs may be reversed on the Secondary server under these circumstances. If this happens, it can be hard to reconcile the names of the NICs with their physical identities, making it difficult to assign the correct IP address to each NIC on the Secondary server.
Resolution
It is part of the normal Neverfail Engine installation process to manually assign the correct IP addresses to each NIC on the Secondary server. If there is no channel connection between the servers, check that the IP addresses on the Secondary server's channel NICs are correctly configured. You should also double-check the settings for the principal (public) NIC, since any configuration error here may not be apparent until a switchover is performed or a failover occurs.
It is possible to capture the identities of all of the NICs on the Secondary server prior to installing Neverfail Engine, by opening a Windows Command Prompt on that server and executing the following command:
ipconfig /all > ipconfig.txt
This saves the current name, TCP/IP configuration, and MAC address of each NIC on the Secondary server to a file called ipconfig.txt, which will be present on that server after the Plug and Play phase of the Neverfail Enigne install has completed. At this point, it is possible to compare the pre-install and post-install state of each NIC by running 'ipconfig /all' from a Windows command prompt and comparing the output of this command with the content of the file ipconfig.txt. The MAC address of each NIC is tied to the physical identity of each card, and never changes - so it is possible to identify each NIC by its MAC address and determine its original name and network configuration, even if these have been updated by the Plug and Play process.
7. Incorrect Connection Selected in the Public Page of the Configure Server UI
Symptoms
Cause
Configuration of the Public page of the Configure Server wizard is incorrect. The Neverfail Channel connection was selected and appears in the NIC field of the Public page in the Configure Server wizard.
Resolution
Reconfigure the Public page of the Configure Server wizard so that the NIC field contains the Public connection.
8. Routing issues:
a. In a LAN
Symptoms
The Neverfail Channel disconnects or fails to connect in a LAN deployment.
Cause
The Neverfail Channel may disconnect or fail to connect due to the principal (public) NIC and/or one or more channels sharing the same subnet
Resolution
If Neverfail Engine is deployed in a LAN environment, the principal (public) IP address and the channel IP address on a server should be in separate subnets. If there are multiple redundant channels, each should have its own subnet. Check the network configuration for each NIC on both servers in the pair, and correct any issues.
Note: If it is not possible to use different subnets for the Public and Channel, static routes between the two channel connections might have to be configured.
b. In a WAN
Symptoms
The Neverfail Channel disconnects or fails to connect in a WAN deployment.
Cause
When the Neverfail Channel disconnects or fails to connect in a WAN deployment it may be the result of the static route not being configured or that it has been configured incorrectly.
When Neverfail Engine is deployed in a WAN, it is generally not possible for the principal (public) IP address and the channel IP addresses to be in different subnets, since there is usually a single network path between the two servers. In order to ensure that channel traffic is routed only between the endpoints of the channel, it is necessary to configure a static route between these endpoints.
Resolution
Please refer to Knowledge Base article #466 - 'How to create a static route for the Neverfail Channel connection in a WAN environment', for a detailed discussion about WAN channel routing issues, and for instructions on how to configure a static route for the Neverfail Channel.
9. Clock/Time settings
Symptoms
The Neverfail Channel disconnects (apparently) at random, but reconnects soon after.
Cause
The time is changed on one of the servers; the Windows Event Log might contain an event detailing by how much the time has been changed:
"The system time has changed to <new time stamp> from <old time stamp>"
The time change will cause the server to incorrectly assume that it has not received any response from the other server and report a "Channel Disconnected"event.
Resolution
All Versions
None
KBID-2855