Snapshotting malfunctioning

[lucventurini]

Whenever I try to have ktserver create a snapshot after its run, ktserver completes normaly but does not create a snapshot. When I delved a bit further, and modified a bit the code to provide the error message, I got this:

2017-11-29T12:57:28.292423Z: [ERROR]: database error: 8: logical inconsistency: hanging

Is there any particular reason this should be happening? I have tried with multiple versions from this repository and the CloudFlare version, to no avail.

Command lines are as follows:

$ ktserver -port 27732 -ls -tout 200000 -th 64 -bgs /tmp/kt/snapshot -bgsc lzo -bgsi 1000000 -log /tmp/kt/log :#opts=ls#bnum=30m#msiz=50g#ktopts=p &
$ ktremotemgr set -host localhost -port 27732 japan tokyo
$ ktremotemgr get -host localhost -port 27732 japan
tokyo
$ kill -SIGINT $(pgrep ktserver)

Thank you for your kind help.

[carlosefr]

How are you telling ktserver to do a snapshot? From your configuration it will only create a snapshot 11 days after being started (-bgsi 1000000). But it should write one on exit, right.

Have you tried removing those parameters one at a time to see which one may be causing the issue? Or terminating ktserver with SIGTERM instead?

[lucventurini]

Hello,
sorry for the late reply. ktserver is used inside a program (cactus) to keep temporary information in memory and save it to disk only at the end of a particular step. Hence the long bgsi value.

I found the trigger of the problem, which is, the fact that the database was initialised in RAM rather than on the disk (":" at the beginning of the parameters section). So for example:

$ ktserver -port 27732 -ls -tout 200000 -th 64 -bgs /tmp/kt/snapshot -bgsc lzo -bgsi 1000000 -log /tmp/kt/log :#opts=ls#bnum=30m#msiz=50g#ktopts=p

will fail to snapshot, but

$ ktserver -port 27732 -ls -tout 200000 -th 64 -bgs /tmp/kt/snapshot -bgsc lzo -bgsi 1000000 -log /tmp/kt/log /tmp/kt.iwch#opts=ls#bnum=30m#msiz=50g#ktopts=p

will. I am still unsure about why this is happening, though.