Hi, could it be that the job has restarted due to failures a large number of times?
Cheers, Aljoscha On Mon, 7 Nov 2016 at 09:21 Dominique Rondé <dominique.ro...@allsecur.de> wrote: > First of all, thanks for the explanation. That sounds reasonable. > > But I started the flink routes 3 days ago and went out for the weekend. > Since we are two people with access to flink i guess there is something > strange happen... > > Greets > Dominique > > > > Von meinem Samsung Gerät gesendet. > > > -------- Ursprüngliche Nachricht -------- > Von: "xiaogang.sxg" <xiaogang....@alibaba-inc.com> > Datum: 07.11.16 04:00 (GMT+01:00) > An: user@flink.apache.org > Betreff: Re: Lot of RocksDB files in tmp Directory > > RocksDB is orginally implemented in C. To use it in Java, Flink uses a > RocksDB jar with JNI techniques. > > Whenever a Flink job starts, the RocksDB libraries will be loaded. The > file named librocksdb*.so will be extracted from the RocksDB jar and will > be put in the tmp directory. So I think you have run Flink jobs many many > times in the three days to get those 451 files. > > I don’t think the numerous files are a big problem because they are > located in the tmp directory and they can be automatically cleaned by OS. > > 在 2016年11月7日，上午8:24，Dominique Rondé <dominique.ro...@allsecur.de> 写道： > > Hi @ll, > > we just change the backend from filesystem to RocksDB. Since that time (3 > Days) we got 451 files with 1.8 GB stored in the tmp-Directory. All files > are named librocksdb*.so > > Did we something wrong or is it a Bug? > > Greets > Dominique > > > > Von meinem Samsung Gerät gesendet. > > >
