Eugene Lazin

OK, I'll take a look

`apr_dbd_sqlite3.so` depend on apr library and doesn't load it explicitly so the executable that loads it should load both. ```bash export LD_PRELOAD="/usr/lib/x86_64-linux-gnu/apr-util-1/apr_dbd_sqlite3.so /usr/lib/x86_64-linux-gnu/libapr-1.so" ```

I think that apr-util-dbd loads apr_dbd_sqlite3 dynamically using the dlopen call. Maybe I can add apr_dbd_sqlite3 as an implicit dependency, let me try.

I added workaround for this. Not sure that it will work though.

The libakumuli.so already links this two libraries so everything should work without LD_PRELOAD.

Thanks for reporting this, I'll take a closer look.

I'll take a look. Might be system cache behavior. It might prefetch part of the data volume into memory. BTW, what query are you using in Grafana?

This can happen if you haven’t cleared the database between the runs of the ingestion tool from the influxdb-comparisons. It'll try to write the same data twice. And, I believe...

You should install libstdc++ and use it with clang (pass -stdlib=libc++ option to the compiler). There is no clang support in my cmake scripts but you can easily add this...

Experiment: - run the akumulid server - run the script with parameters: `--tbegin=20170101T000000 --tend=20170101T001000 --tdelta=00:00:10 --rbegin=0 --rend=100000` (create 1M new series) - restart the akumulid server - run the script...