|
1 | 1 |
|
2 | 2 | a| https://jira.issues.couchbase.com/browse/MB-67762[MB-67762] |
3 | 3 |
|
4 | | -a| We have identified an integer-overflow bug for buckets using the Magma storage engine. |
5 | | - |
6 | | -This bug can, under specific conditions, cause a subset of documents to become invisible to read operations. |
7 | | -A Magma Bucket becomes susceptible after approximately 500 billion to 2 trillion mutations per bucket (500 M to 2B per vBucket partition). |
8 | | -It is important to note that this counter is reset whenever the vBucket is rebalanced, |
9 | | -effectively resetting the timeline toward this threshold. |
10 | | - |
11 | | -In addition, there is also no impact if all the documents are touched within the 500 billion to 2 trillion threshold, |
12 | | -as in the case when expiry has been triggered on all the documents before the bucket becomes susceptible. |
13 | | - |
14 | | -Otherwise, under specific conditions, this issue can lead to potential long-term data integrity issues if the bucket is not rebalanced before accumulating 500 billion to 2 trillion mutations. |
15 | | -There is also no impact if all the documents are touched within the 500 billion to 2 trillion threshold, for example, expiry has been triggered on |
16 | | - |
17 | | - |
18 | | -a| |
19 | | -For clusters identified as being vulnerable, use the following steps to resolve the issues: |
20 | | - |
21 | | -[discrete] |
22 | | -===== {counter: seq}. Stop Creating the new index |
23 | | -Stop building any index while applying the resolution steps. |
24 | | - |
25 | | -[discrete] |
26 | | -===== {counter: seq}. Run Bucket Compaction |
27 | | - |
28 | | -Compaction needs to be run for the couchbase buckets which are exposed to the bug as reported by the diagnostic script. |
29 | | - |
30 | | -[discrete] |
31 | | -===== {counter: seq}. Upgrade Couchbase Server |
32 | | -Upgrade the Couchbase Server on all nodes. You can perform the upgrade using one of two methods ONLY: |
33 | | - |
34 | | -* A graceful failover of each node |
35 | | -* An offline upgrade of the Couchbase Server |
36 | | - |
37 | | -[discrete] |
38 | | -===== {counter: seq}. Backup Data |
39 | | -Backup Magma data files for nodes with any vulnerable vBucket after gracefully failing over the node or taking the node offline. |
40 | | -To back up Magma data files, copy the /data/bucket directory to a backup location. Note that the copy should be performed after stopping couchbase server to maintain the consistency of the data files: |
41 | | - |
42 | | -For clusters identified as being vulnerable, use the following steps to resolve the issues. |
43 | | - |
44 | | -[discrete] |
45 | | -===== {counter: seq}. Stop Creating New index |
46 | | - |
47 | | -Stop building any index while applying the resolution steps. |
48 | | - |
49 | | -[discrete] |
50 | | -===== {counter: seq}. Run Bucket Compaction |
51 | | - |
52 | | -Compaction needs to be run for the couchbase buckets which are exposed to the bug as reported by the diagnostic script. |
53 | | - |
54 | | -[discrete] |
55 | | -===== {counter: seq}. Upgrade Couchbase Server |
56 | | - |
57 | | -Upgrade the Couchbase Server on all nodes. You can perform the upgrade using one of two methods ONLY: |
58 | | - |
59 | | -* A graceful failover of each node |
60 | | -* An offline upgrade of the Couchbase Server |
61 | | - |
62 | | -[discrete] |
63 | | -===== {counter: seq}. Backup Data |
64 | | - |
65 | | -Backup Magma data files for nodes with any vulnerable vBucket after gracefully failing over the node or taking the node offline. To back up Magma data files, copy the /data/bucket directory to a backup location. Note that the copy should be performed after stopping couchbase server to maintain the consistency of the data files: |
66 | | - |
67 | | -[source,shellscript] |
68 | | ----- |
69 | | -# service couchbase-server stop |
70 | | -
|
71 | | -# cp -rp /data/bucketname /backup_data/ |
72 | | ----- |
73 | | - |
74 | | -[discrete] |
75 | | -===== {counter: seq}. Perform Delta Recovery and Restart Couchbase Server |
76 | | - |
77 | | -Once this step is complete, any documents previously hidden in the vulnerable window will immediately become accessible. |
78 | | - |
79 | | -[discrete] |
80 | | -===== {counter: seq}. Reclaim Wasted Disk Space |
81 | | - |
82 | | -This step addresses potential storage bloat caused by old document versions not being garbage-collected. To reclaim this disk space, you can perform swap rebalance on the _nodes that the diagnostic script identified_ as being impacted. |
83 | | -== Perform Delta Recovery and Restart Couchbase Server |
84 | | - |
85 | | -Once this step is complete, any documents previously hidden in the vulnerable window will immediately become accessible. |
| 4 | +a| There is an integer overflow bug in buckets using the Magma storage engine. |
86 | 5 |
|
| 6 | +a| For details, please refer to https://jira.issues.couchbase.com/browse/MB-67762[MB-67762] |
0 commit comments