"
=== How the the partition level repair works
- The repair master decides which ranges to work on.
- The repair master splits the ranges to sub ranges which contains around 100
partitions.
- The repair master computes the checksum of the 100 partitions and asks the
related peers to compute the checksum of the 100 partitions.
- If the checksum matches, the data in this sub range is synced.
- If the checksum mismatches, repair master fetches the data from all the peers
and sends back the merged data to peers.
=== Major problems with partition level repair
- A mismatch of a single row in any of the 100 partitions causes 100
partitions to be transferred. A single partition can be very large. Not to
mention the size of 100 partitions.
- Checksum (find the mismatch) and streaming (fix the mismatch) will read the
same data twice
=== Row level repair
Row level checksum and synchronization: detect row level mismatch and transfer
only the mismatch
=== How the row level repair works
- To solve the problem of reading data twice
Read the data only once for both checksum and synchronization between nodes.
We work on a small range which contains only a few mega bytes of rows,
We read all the rows within the small range into memory. Find the
mismatch and send the mismatch rows between peers.
We need to find a sync boundary among the nodes which contains only N bytes of
rows.
- To solve the problem of sending unnecessary data.
We need to find the mismatched rows between nodes and only send the delta.
The problem is called set reconciliation problem which is a common problem in
distributed systems.
For example:
Node1 has set1 = {row1, row2, row3}
Node2 has set2 = { row2, row3}
Node3 has set3 = {row1, row2, row4}
To repair:
Node1 fetches nothing from Node2 (set2 - set1), fetches row4 (set3 - set1) from Node3.
Node1 sends row1 and row4 (set1 + set2 + set3 - set2) to Node2
Node1 sends row3 (set1 + set2 + set3 - set3) to Node3.
=== How to implement repair with set reconciliation
- Step A: Negotiate sync boundary
class repair_sync_boundary {
dht::decorated_key pk;
position_in_partition position
}
Reads rows from disk into row buffers until the size is larger than N
bytes. Return the repair_sync_boundary of the last mutation_fragment we
read from disk. The smallest repair_sync_boundary of all nodes is
set as the current_sync_boundary.
- Step B: Get missing rows from peer nodes so that repair master contains all the rows
Request combined hashes from all nodes between last_sync_boundary and
current_sync_boundary. If the combined hashes from all nodes are identical,
data is synced, goto Step A. If not, request the full hashes from peers.
At this point, the repair master knows exactly what rows are missing. Request the
missing rows from peer nodes.
Now, local node contains all the rows.
- Step C: Send missing rows to the peer nodes
Since local node also knows what peer nodes own, it sends the missing rows to
the peer nodes.
=== How the RPC API looks like
- repair_range_start()
Step A:
- request_sync_boundary()
Step B:
- request_combined_row_hashes()
- reqeust_full_row_hashes()
- request_row_diff()
Step C:
- send_row_diff()
- repair_range_stop()
=== Performance evaluation
We created a cluster of 3 Scylla nodes on AWS using i3.xlarge instance. We
created a keyspace with a replication factor of 3 and inserted 1 billion
rows to each of the 3 nodes. Each node has 241 GiB of data.
We tested 3 cases below.
1) 0% synced: one of the node has zero data. The other two nodes have 1 billion identical rows.
Time to repair:
old = 87 min
new = 70 min (rebuild took 50 minutes)
improvement = 19.54%
2) 100% synced: all of the 3 nodes have 1 billion identical rows.
Time to repair:
old = 43 min
new = 24 min
improvement = 44.18%
3) 99.9% synced: each node has 1 billion identical rows and 1 billion * 0.1% distinct rows.
Time to repair:
old: 211 min
new: 44 min
improvement: 79.15%
Bytes sent on wire for repair:
old: tx= 162 GiB, rx = 90 GiB
new: tx= 1.15 GiB, tx = 0.57 GiB
improvement: tx = 99.29%, rx = 99.36%
It is worth noting that row level repair sends and receives exactly the
number of rows needed in theory.
In this test case, repair master needs to receives 2 million rows and
sends 4 million rows. Here are the details: Each node has 1 billion *
0.1% distinct rows, that is 1 million rows. So repair master receives 1
million rows from repair slave 1 and 1 million rows from repair slave 2.
Repair master sends 1 million rows from repair master and 1 million rows
received from repair slave 1 to repair slave 2. Repair master sends
sends 1 million rows from repair master and 1 million rows received from
repair slave 2 to repair slave 1.
In the result, we saw the rows on wire were as expected.
tx_row_nr = 1000505 + 999619 + 1001257 + 998619 (4 shards, the numbers are for each shard) = 4'000'000
rx_row_nr = 500233 + 500235 + 499559 + 499973 (4 shards, the numbers are for each shard) = 2'000'000
Fixes: #3033
Tests: dtests/repair_additional_test.py
"
* 'asias/row_level_repair_v7' of github.com:cloudius-systems/seastar-dev: (51 commits)
repair: Enable row level repair
repair: Add row_level_repair
repair: Add docs for row level repair
repair: Add repair_init_messaging_service_handler
repair: Add repair_meta
repair: Add repair_writer
repair: Add repair_reader
repair: Add repair_row
repair: Add fragment_hasher
repair: Add decorated_key_with_hash
repair: Add get_random_seed
repair: Add get_common_diff_detect_algorithm
repair: Add shard_config
repair: Add suportted_diff_detect_algorithms
repair: Add repair_stats to repair_info
repair: Introduce repair_stats
flat_mutation_reader: Add make_generating_reader
storage_service: Introduce ROW_LEVEL_REPAIR feature
messaging_service: Add RPC verbs for row level repair
repair: Export the repair logger
...
c96fc1d585