Filipe Regadas

Add Memcached IO

1

`Memcached` is widely used for a lot of use cases. It would be nice if we could support it more seamlessly.

Reevaluate Zoltar in scio-tensorflow

1

We are only leveraging Zoltar for model loading. Since we are not leveraging the other features maybe we can live without it. That said, let's see if we can remove...

Currently we have `scio-cassandra3` should we have a `scio-cassandra4`; There are some client API changes but overall seems that what we have can:tm: work

Flink execution can throw EOFException on serialization

1

I haven't looked into this yet, but I'm seeing these popping up from time to time. ``` java.io.EOFException: Can't collect further: memorySource depleted at Due to Exception while trying to...

Scio streaming pipeline stuck in Bigtable lookup

3

``` Operation ongoing in step Read from Master-Index for at least 05m00s without outputting or completing in state finish at sun.misc.Unsafe.park(Native Method) at java.util.concurrent.locks.LockSupport.parkNanos(LockSupport.java:215) at com.google.common.util.concurrent.AbstractFuture.get(AbstractFuture.java:412) at com.google.common.util.concurrent.AbstractFuture$TrustedFuture.get(AbstractFuture.java:90) at com.foo.BigTableLookupDoFn.waitForFutures(BigTableLookupDoFn.scala:36)...

I'm noticing this while using the `ScalaAsyncLookupDoFn`

This issue happens on serde when Java `serialVersionUID` is checked. Scala 2.12.x doesn't extensively set `@SerialVersionUID` across collections which means that in order for this to work in Flink one...

Decouple algebird ops from scio-core

2

At some point i think it would be good to have a scio-algebird specific module. It might even ease up the migration for 2.13.