tests.md

TODOs

1. Partitioning - Done
2. Performance and throughput functions
3. PrintDataStore function - Done
4. Support float for amt - Done
5. Fix transactions with amt = balance
6. Elections are overlapping a lot - Done
7. print_committed_txns prints transactions aborted due to 2PC but present in txt files

Tests

1. Concurrent intra-shard with independent items in single cluster

client 1: 1,2,1 client 2: 4,5,1 client 3: 7,8,1

Output 1 - 9 2 - 11 4 - 9 5 - 11 7 - 9 8 - 11

Concurrent intra-shard with independent items in different clusters

client 1: 100,101,1 client 2: 1100,1101,1 client 3: 2200,2201,1

Output 100 - 9 101 - 11 1100 - 9 1101 - 11 2200 - 9 2201 - 11

Concurrent intra-shard with same items in single cluster

client 1: 1500,1501,1 client 2: 1500,1502,1 client 3: 1503,1500,1

Output 1500 - 9 1501 - 11 1502 - 11 1503 - 9

Simple cross-shard transaction

client 1: 10,1010,1

Output 10 - 9 1010 - 11

Concurrent cross-shard transactions on different items

client 1: 200,1200,1 client 2: 1300,2300,1 client 3: 2500,500,1

Output 200 - 9 1200 - 11 1300 - 9 2300 - 11 2500 - 9 500 - 11

Concurrent cross-shard transactions on same items

client 1: 20,1020,1 client 2: 1020,30,1 client 3: 30,2020,1

Output If 1 and 2 get aborted 20 - 10 30 - 9 1020 - 10 2020 - 11

Concurrent cross-shard transactions on same items

client 1: 60,1600,1 client 2: 1600,66,1 client 3: 1800,55,1

Output If request 2 gets aborted 55 - 11 60 - 9 66 - 10 1600 - 11 1800 - 9

Concurrent intra-shard and cross-shard accessing different items

Client 1: 11,22,1 Client 2: 1222,2333,1 Client 3: 33,44,1

Concurrent intra-shard and cross-shard accessing same items

Client 1: 1,2,1 (press enter first) Client 2: 1500,2,1 (press enter second) Output If 1500,2,1 fails 1 - 9 2 - 11 1500 - 10

Client 1: 1,2,1 (press enter second) Client 2: 2,1500,1 (press enter first) Output If both go through 1 - (9-1) = 8 2 - (11-1+1) = 11 1500 - 11

Leader fails and then initiate concurrent transactions (intra-shard and cross-shard)

Client 1: 3,4,1 Client 2: 3,1300,1 Output If second txn gets aborted 3 - 9 4 - 11 1300 - 10

2 nodes fail and then initiate intra-shard

Output No progress is made. Once nodes come back alive, transaction resumes

2 nodes fail and initiate cross-shard

Kill server 2,3, ensure 1 is the leader Client 1: 11,1200,1 Output 1 should make no progress and wait until majority to send a YES/NO to client Once 2 is up, it sends a YES for 2PC and executes txn. 3 also gets updated

Concurrent intra-shard transactions on same items with

Partition leader + 2 nodes and intra-shard transaction

If server 3 is leader and it gets partitioned, send 1,2,1 from client Output If request goes to partitioned leader, client retries and works in the majority nodes to make progress. 1 - 9 2 - 11

Partition leader + 2 nodes on both clusters and cross-shard

Output A new leader should get elected and make progress. If the request goes to the old leader, it should timeout and the client will retry eventually reaching the group of 2 nodes.

Partition leader with a node + another node and intra-shard transaction

Output Another node which is partitioned continuously tries to become leader but does not get enough votes. Leader and other node make progress.

Intra-shard with amt > balance

Output Aborts transaction

Cross-shard with amt > balance for 1 of them

Output Aborts transaction

Intra-shard with amt = balance

Output Works as expected

Cross-shard with amt = balance

Output Works as expected

Partition off non-leader and ensure request going to partitioned node is retried

Output Works fine

Recovery from partitioning after partitioning leader

Recovery from partitioning after partitioning non-leader node