
F5 SDC Troubleshooting Guide
Proprietary and Confidential Information of F5 Networks
21
Causes
All pools are in “Out of Service” state (since all peers of each of these pools are
in a “Close” state (disabled or not yet connected) or “Out of Service”).
At least one pool is in an “Open” state – but all of its peers are overloaded
(reached maximum rate limit), and all other pools, if exist, are in “Out of
Service” state.
Note: A pool will be in “Out of Service” state when at least its “Minimum Number
of Peers” (configurable, default is 1) is reached. That means that there are no
“Minimum Number of Peers” peers in this pool such that their state is “Open”.
Symptoms
The following INFO message may appear in the CPF logs:
“Unable to choose pool: {0}, reason: {1}” for each of the pools that belong to
the selected routing row.
The following WARN message may appear in the CPF logs:
“Failed to select a Pool to handle a request received from {0}. The selected
routing row index is {1} with policy {2}, the incoming message is {3}”.
Resolution
Check the condition of the peers of each pool of the selected routing row to
make sure they are not overloaded/closed/disabled, etc.
Change the logging of TransactionManagement to DEBUG and follow the CPF
logs to trace the pool selection.
5.2.3 Endless Pending Request Timeouts toward Client
Error Description
When CPF reaches a state where it cannot route a request to a server, it sends an error answer
to the client. This error answer can be edited for each row of the routing table using the
“Handle Server Error” scripts. Faulty scripts can cause SDC to behave very strangely. For
example, if a script returns answerFromServer and the RemoteNodeEvent is
CANNOT_ROUTE, then the message that is sent to client is the request that was sent by it.
Since it is a request that goes downstream, now CPF will set it as a pending request with a
Commentaires sur ces manuels