When resending a large number of messages, I’ll frequently just run a script from a Tcl shell.
I’ll loop through messages one at a time – write each message to a temporary file and queue messages with hcicmd using a command something like this:
if { [catch {exec hcicmd -p $process -c “$thread resend ib_pre_tps data 5120 /tmp/requeueFile.txt nl”} err] } { puts “ERROR: $err” }
You need to setsite to the proper site prior to queueing the messages.
To throttle the message flow, I’ll insert a sleep for a variable number of milliseconds between each message. The number of msec is read from a file.
I keep track of the last mod time of that file and compare it to current mod time of the file. When the mod time changes, I update the sleep msec between messages. Something like this:
if { [file exists $sleepFilePath] } {
set sleepFileMtime [file mtime $sleepFilePath]
if { $sleepFileMtime > $lastSleepMtime } {
set sf [open $sleepFilePath r]
set tmpSleepMsec [string trim [read $sf]]
close $sf
if { [string is digit $tmpSleepMsec] } {
puts “setting sleepMsec to $tmpSleepMsec”
set sleepMsec $tmpSleepMsec
}
set lastSleepMtime $sleepFileMtime
}
} else {
set sf [open $sleepFilePath w]
puts $sf $sleepMsec
close $sf
set lastSleepMtime [file mtime $sleepFilePath]
}
This allows you to fine-tune the resend rate on-the-fly.
Start off conservative with a second (1000 msec) between messages. Monitor recovery DB queue depth and adjust the sleep time accordingly. Reduce it to send messages faster. Increase it if the queue grows.
As soon as you write a new value and save the file, the script will see the mod time has changed, read the new value from the file and adjust the sleep msec.
After a short tuning time, you’ll arrive at a number that sends the messages as quickly as possible, but doesn’t result in a recovery DB that grows too large.
Jeff Dinsmore
Chesapeake Regional Healthcare
.
