export | import doesnt load data, only creates table

[keen99]

ok, I've been banging my head on this too long now.

Simple summary - import (and put) dont actually load data into ddb.

I've also tested this with put (removing the first line that contains the table schema from the export) after creating a table, with the same result. empty stdout/stderr, and no data added to ddb.

running v1.4.1 code:

%% dyno export us-west-2/dsr-test
{"AttributeDefinitions":[{"AttributeName":"user","AttributeType":"S"}],"TableName":"dsr-test","KeySchema":[{"AttributeName":"user","KeyType":"HASH"}],"ProvisionedThroughput":{"ReadCapacityUnits":8,"WriteCapacityUnits":2},"TableArn":"arn:aws:dynamodb:us-west-2:569641690693:table/dsr-test"}
{"user":{"S":"testser"}}
{"user":{"S":"teetst"}}

%% dyno export us-west-2/dsr-test | dyno import us-west-2/dsr-test-restore-1

%% dyno export us-west-2/dsr-test-restore-1
{"AttributeDefinitions":[{"AttributeName":"user","AttributeType":"S"}],"TableName":"dsr-test-restore-1","KeySchema":[{"AttributeName":"user","KeyType":"HASH"}],"ProvisionedThroughput":{"ReadCapacityUnits":8,"WriteCapacityUnits":2},"TableArn":"arn:aws:dynamodb:us-west-2:569641690693:table/dsr-test-restore-1"}

%%

The tests added in #117 for #96 seem to pass:

cheese:~/git/github/dyno-clone((v1.4.1))%% npm uninstall
cheese:~/git/github/dyno-clone((v1.4.1))%% npm install
cheese:~/git/github/dyno-clone((v1.4.1))%% export PATH="$(npm bin):$PATH"
cheese:~/git/github/dyno-clone((v1.4.1))%% node test/cli.test.js
...
# [cli] export --> import roundtrip
ok 31 cli success
ok 32 cli success
ok 33 loaded records into table

and using the direct code (vs the npm installed package) has the same result:

cheese:~/git/github/dyno-clone((v1.4.1))%% node bin/cli.js export us-west-2/dsr-test
{"AttributeDefinitions":[{"AttributeName":"user","AttributeType":"S"}],"TableName":"dsr-test","KeySchema":[{"AttributeName":"user","KeyType":"HASH"}],"ProvisionedThroughput":{"ReadCapacityUnits":8,"WriteCapacityUnits":2},"TableArn":"arn:aws:dynamodb:us-west-2:569641690693:table/dsr-test"}
{"user":{"S":"testser"}}
{"user":{"S":"teetst"}}

cheese:~/git/github/dyno-clone((v1.4.1))%% node bin/cli.js export us-west-2/dsr-test | node bin/cli.js import us-west-2/dsr-test-restore-3

cheese:~/git/github/dyno-clone((v1.4.1))%% node bin/cli.js export us-west-2/dsr-test-restore-3
{"AttributeDefinitions":[{"AttributeName":"user","AttributeType":"S"}],"TableName":"dsr-test-restore-3","KeySchema":[{"AttributeName":"user","KeyType":"HASH"}],"ProvisionedThroughput":{"ReadCapacityUnits":8,"WriteCapacityUnits":2},"TableArn":"arn:aws:dynamodb:us-west-2:569641690693:table/dsr-test-restore-3"}

cheese:~/git/github/dyno-clone((v1.4.1))%%

[mcwhittemore]

@keen99 did you make any progress on this? I haven't used this feature before.

/cc @rclark

[keen99]

not really - all I managed to confirm through hackery was that the first line is all that ever gets read in import, and in put the lines get read but Importer is never called. (roughly. it's been a long long long few days since I was in there.)

[rclark]

Yeah there's definitely at least a few things wrong:

• It's possible that the pause/resume logic flow here doesn't work like it was intended to. This could cause the data records in an import operation to "vanish".

• DynamoDB Stream information needs to be removed from export output prior to an import command. This ought to be done in the cleanDescription function

• The Parser() is doing some kinda screwball something to the first line in the data it receives. Firing the callback prior to the return here makes the dyno put work the way it should, but still didn't get information across on export/import.

[keen99]

thanks! firing the callback before Parser's firstline returns got put working for me (as long as the first line isn't schema)

    if (firstline) {
      firstline = false;

      var parsed = Dyno.deserialize(record);
      setImmediate(callback);
      if (!Object.keys(parsed).every(function(key) {
        return !!parsed[key];
      })) return this.push(JSON.parse(record.toString()));
    }

so with this I can at least create from a schema definition and load data - since I already have a custom "export" wrapper tool that does just schema dumps (mapbox/dynamodb-replicator#85), I'll probably just add modified versions of the import/put tools to that to get past my restore hurdles for dynamodb-replicator for now.

thanks!

[keen99]

well, looks like put in the above case only works for 2550 lines of input - on 2551 (with some light variation testing, it seems to always be the 2551th line)

cheese:~/git/github/dynamodb-replicator(dsr-hackery)%% cat $TABLE.out.json | tail -n2551 | node bin/dyno.js put $AWS_REGION/dsr-test-restore-$TABLE
Error: no writecb in Transform class
    at afterTransform (_stream_transform.js:71:33)
    at TransformState.afterTransform (_stream_transform.js:54:12)
    at Transform.Importer.importer._transform (/Users/draistrick/git/github/dynamodb-replicator/bin/dyno.js:211:7)
    at runCallback (timers.js:576:20)
    at tryOnImmediate (timers.js:550:5)
    at processImmediate [as _immediateCallback] (timers.js:529:5)
cheese:~/git/github/dynamodb-replicator(dsr-hackery)%% cat $TABLE.out.json | tail -n2550 | node bin/dyno.js put $AWS_REGION/dsr-test-restore-$TABLE
cheese:~/git/github/dynamodb-replicator(dsr-hackery)%%