Hi Anand, In general, transfer size is roughly equal to actual stored size with the addition of file header and entry descriptors. That being said, there's some overhead per stored RR (~16B per record, or so CC @jan.kadlec) since the RRs aren't merged into RRSets to keep ordering and such. But as a rule of thumb, you should allow at least twice the desired maximum transfer size for a journal size for two reasons: - the journal will be able to keep history of at least one transfer back - won't need to flush the zone file after each change (helps to avoid disk usage spikes) Best, Marek On 15 December 2014 at 11:16, Anand Buddhdev <anandb(a)ripe.net> wrote: > Hi Knot developers, > > I've now installed version 1.6.1 on some servers, and I'm observing some > journal related issues, and I have questions about them. First off, > here's one issue: > > 2014-12-15T06:00:56 info: [203.in-addr.arpa] NOTIFY, incoming, > 193.0.0.198@53535: received serial 3006121318 > 2014-12-15T06:00:56 info: [203.in-addr.arpa] refresh, outgoing, > 193.0.0.198@53: master has newer serial 3006121317 -> 3006121318 > 2014-12-15T06:00:56 info: [203.in-addr.arpa] IXFR, incoming, > 193.0.0.198@53: starting > 2014-12-15T06:00:57 warning: [203.in-addr.arpa] IXFR, incoming, > 193.0.0.198@53: failed to write changes to journal (not enough space > provided) > 2014-12-15T06:00:58 notice: [203.in-addr.arpa] IXFR, incoming, > 193.0.0.198@53: fallback to AXFR > 2014-12-15T06:00:58 info: [203.in-addr.arpa] AXFR, incoming, > 193.0.0.198@53: starting > 2014-12-15T06:00:59 info: [203.in-addr.arpa] AXFR, incoming, > 193.0.0.198@53: finished, serial 3006121317 -> 3006121318, 0.65 seconds, > 171 messages, 7960312 bytes > > So it looks like the IXFR is too big, and won't fit into the journal, > and Knot is falling back to AXFR. When I requested this IXFR by hand, I got: > > $ dig ixfr=3006121317 203.in-addr.arpa @193.0.0.198 > ... > ... > ;; XFR size: 121779 records (messages 170, bytes 7963389) > > The size of the IXFR in bytes is below the configured file size limit > (10M), but I suspect that 7963389 bytes probably take up more room in > the journal, so Knot can't write into it, and is falling back to AXFR. > > Are you able to tell me (approximately of course), how much disk space > is required for a given number of bytes of IXFR? This will help me tune > the setting of ixfr-fslimit to avoid this unnecessary fallback to AXFR. > _______________________________________________ > knot-dns-users mailing list > knot-dns-users(a)lists.nic.cz > https://lists.nic.cz/cgi-bin/mailman/listinfo/knot-dns-users