out of memory handling during *XFR
paul at xelerance.com
Thu Aug 20 22:05:47 UTC 2009
May I suggest an improvement in the handling of bind's out of memory
handling when performing *XFR's?
I am talking about these: failed while receiving responses: out of memory
Currently, bind drops the AXFR, and I assume the memory of the failed
partial *XFR'd zone, and tries again. On dedicated name servers, where
memory does not spontaniously appears from no where, this just starts
a loop of endless *XFR requests, peaking out a lot of bandwidth.
Bind could check to see if it got more free memory at the start of the
run then it had at the previously failed run.
Bind could do an exponential back off for the *XFR's.
More information about the bind-users