6100 ADCCP Programming Manual

Recovery Strategies
File-System Error Codes
A–2 069225 Tandem Computers Incorporated
Case 3 You were establishing or severing a connection when the error occurred. The
best way to handle a mode-setting, line-bidding, or connection sequence is to
abort the sequence and start again from the beginning.
You may need to try your first WRITEREAD several times, until the line is in a state in
which the protocol can accept the request. For example, if the failure occurs while the
protocol is waiting to read, it may still be monitoring the line for data when you retry;
in that case, your application receives a message such as “wrong line state for this
operation,” and your application must keep trying until the earlier read operation
times out.
The request ID of a retry must match the request ID of the original request. The
request ID keeps the protocol from sending the same data twice or from delivering the
same data twice to the application.
The behavior of another station can also affect recovery. For instance, if the remote
station notes a delay and resets the line, its action causes a change in state, requiring
end-to-end recovery. CP6100 fault-tolerance cannot extend beyond the subsystem
cabinet.
Total Subsystem and LIU
Failures
An error condition sometimes affects more than one path to a device. CP6100 attempts
to recover from failures affecting more than one path to a device without adversely
affecting running applications.
If a path does not exist between the controllers and the 6100 subsystem cabinet (for
instance, if both cables are unplugged or the subsystem cabinet has no power), CP6100
queues user requests and completes them every 30 seconds with file system error 231.
When power returns, CP6100 downloads the line. Applications can resume using the
line, with the following restriction:
If the SYSGEN AUTOCLOSE parameter is set, applications must close and reopen
the line to use it again.
If the SYSGEN AUTOCONF parameter is set and applications have made
configuration changes, those changes must be restored (because the configuration
block has been overwritten).
The delay resulting from the failure may have caused a timeout or loss of the
communication link. It may be necessary to perform end-to-end recovery.
These same restrictions apply whenever the line is downloaded as the result of a CMI
START command or as part of an error recovery.
If an LIU does not respond to a request, CP6100 tries to establish contact through the
Communications Subsystem Manager (CSM) a maximum of three times. In the
meantime, requests finish with error 124. If the LIU fails to respond after three
attempts at recovery, CP6100 declares the line down and completes new requests with
error 66. (An operator must then use CMI to download the line and return it to
service.) If the LIU recovers, applications can resume using the line.