I upgraded recently from 1.7.3 to 1.8.0, since then I'm observing a constant
and steady growth of the requestlist queue. Currently it is about x8
what it usually used to be and it still keeps growing.
Is that expected with some of the new default?
I specifically disabled so-reuseport (see other thread).
The number of qps remained unchanged.
I assume that a growing queue size is bad for service quality
if the number of qps is unchanged.
Since I'm currently not observing any negative impact on reply time
I'm also not ruling out that this might be a monitoring issue (using unbound's
munin plugin). Have there been any changes that would affect the requestlist graph?
I’ve always had a steady growth in the request list I always thought that it is normal. I’m not 100% sure but unbound tries again later to get an answer for the requests in the list but backs-off asking the servers with longer intervals over time.
y=you can use “unbound-control dump_requestlist” to so see what is currently in the list. Most likely you will see *.in-addr.arpa. queries and others that will never be answered.
If you want to you can flush them with “unbound-control flush_requestlist”.
I've always had a steady growth in the request list I always thought that it is normal.
based on the munin plugin description:
"The queries that did not hit the cache and need recursion service take up space in the requestlist.
If there are too many queries, first queries get overwritten, and at last resort dropped."
I upgraded recently from 1.7.3 to 1.8.0, since then I'm observing a constant
and steady growth of the requestlist queue. Currently it is about x8
what it usually used to be and it still keeps growing.
This is a previous bugreport and there are already fixes in the code
repository for it. It shouldn't actually change the responses to normal
queries, but these entries are using up some time and space on the
server for a while. The fix should remove them.
This is a previous bugreport and there are already fixes in the code
repository for it. It shouldn't actually change the responses to normal
queries, but these entries are using up some time and space on the
server for a while. The fix should remove them.
thanks for your reply. I've searched the bugzilla bugtracker a bit
but didn't find it, could you point me to it?