On Thu, May 30, 2019 at 5:33 PM Jeffrey Manno <[email protected]> wrote:
> That makes sense. Do you think we should update accumulo-testing? Currently > I think that makes sense to update the defaults in accumulo-testing to not use snappy. > 'cingest createtable' function > utilizes that compression method which is how the issue was discovered. We > could just remove it to use the default or change it to a different > compression method. > > On Thu, May 30, 2019 at 2:31 PM Mike Miller <[email protected]> > wrote: > > > It is random whether the Hadoop developer who did the release built it > with > > Snappy or whatever options. > > > > You will either have to built Hadoop yourself (have fun with that 😁) or > > use a different version. I ran into this issue when we added > Zstandard. I > > put some notes on the issue here if you want to try building Hadoop: > > https://github.com/apache/accumulo/issues/438 > > > > On Thu, May 30, 2019 at 2:03 PM Jeffrey Manno <[email protected]> > > wrote: > > > > > I did some more digging on this. It does appear that snappy is not > being > > > properly built while the rest of the other native libraries are. I am > not > > > completely sure on why this is. Whether its because of installing > through > > > fluo-uno or with this version of hadoop (3.2.0). I did get past this by > > > adding the libsnappy.so.1 file manually. > > > > > > On Tue, May 28, 2019 at 5:36 PM Jeffrey Manno < > [email protected]> > > > wrote: > > > > > > > I have ran into this issue before. I believe it has something to do > > with > > > > the second part Christopher mentioned. > > > > I know installing snappy into the OS only causes more headaches. I > can > > > > take a deeper look into it tomorrow. > > > > > > > > On Tue, May 28, 2019 at 5:31 PM Christopher <[email protected]> > > wrote: > > > > > > > >> You might need to install snappy into your OS. > > > >> On Fedora: 'sudo dnf install snappy' > > > >> On RHEL/CentOS: 'sudo yum install snappy' > > > >> > > > >> It's also possible that the version of Hadoop you're using doesn't > > > >> have its native libraries built, but it's hard to know more without > > > >> seeing more of the error message (like an associated stack trace or > > > >> logger class name). If it is Hadoop not having it's native libraries > > > >> built, I'm not sure how to rebuild Hadoop from source. > > > >> > > > >> On Tue, May 28, 2019 at 3:48 PM Jeffrey Zeiberg <[email protected] > > > > > >> wrote: > > > >> > > > > >> > I am seeing this in my error logs on my Uno instance when I run > the > > > >> > "./bin/cingest ingest" in accumulo-testing. Anyone else seeing > > this? > > > >> > > > > >> > > > > > > > > > >
