You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: CHANGELOG.md
+79-1Lines changed: 79 additions & 1 deletion
Original file line number
Diff line number
Diff line change
@@ -1,6 +1,7 @@
1
1
# Table of Contents
2
2
3
3
4
+
-[2.1.2](#212)
4
5
-[2.1.1](#211)
5
6
-[2.1.0](#210)
6
7
-[2.0.5](#205)
@@ -46,10 +47,66 @@
46
47
-[0.9.9 and prior](#099---20170202)
47
48
48
49
50
+
## [2.1.2]
51
+
52
+
> Released 2020/08/13
53
+
54
+
:white_check_mark:**Update (2020/08/13)**: This release fixed a balancer
55
+
bug that may cause incorrect request payloads to be sent to unrelated
56
+
upstreams during balancer retries, potentially causing responses for
57
+
other requests to be returned. Therefore it is **highly recommended**
58
+
that Kong users running versions `2.1.0` and `2.1.1` to upgrade to this
59
+
version as soon as possible, or apply mitigation from the
60
+
[2.1.0](#210) section below.
61
+
62
+
### Fixes
63
+
64
+
##### Core
65
+
66
+
- Fix a bug that balancer retries causes incorrect requests to be sent to
67
+
subsequent upstream connections of unrelated requests.
68
+
[#6224](https://github.com/Kong/kong/pull/6224)
69
+
- Fix an issue where plugins iterator was being built before setting the
70
+
default workspace id, therefore indexing the plugins under the wrong workspace.
71
+
[#6206](https://github.com/Kong/kong/pull/6206)
72
+
73
+
##### Migrations
74
+
75
+
- Improve reentrancy of Cassandra migrations.
76
+
[#6206](https://github.com/Kong/kong/pull/6206)
77
+
78
+
##### PDK
79
+
80
+
- Make sure the `kong.response.error` PDK function respects gRPC related
81
+
content types.
82
+
[#6214](https://github.com/Kong/kong/pull/6214)
83
+
84
+
49
85
## [2.1.1]
50
86
51
87
> Released 2020/08/05
52
88
89
+
:red_circle:**Post-release note (as of 2020/08/13)**: A faulty behavior
90
+
has been observed with this change. When Kong proxies using the balancer
91
+
and a request to one of the upstream `Target` fails, Kong might send the
92
+
same request to another healthy `Target` in a different request later,
93
+
causing response for the failed request to be returned.
94
+
95
+
This bug could be mitigated temporarily by disabling upstream keepalive pools.
96
+
It can be achieved by either:
97
+
98
+
1. In `kong.conf`, set `upstream_keepalive_pool_size=0`, or
99
+
2. Setting the environment `KONG_UPSTREAM_KEEPALIVE_POOL_SIZE=0` when starting
100
+
Kong with the CLI.
101
+
102
+
Then restart/reload the Kong instance.
103
+
104
+
Thanks Nham Le (@nhamlh) for reporting it in [#6212](https://github.com/Kong/kong/issues/6212).
105
+
106
+
:white_check_mark:**Update (2020/08/13)**: A fix to this regression has been
107
+
released as part of [2.1.2](#212). See the section of the Changelog related to this
108
+
release for more details.
109
+
53
110
### Dependencies
54
111
55
112
- Bump [lua-multipart](https://github.com/Kong/lua-multipart) to `0.5.9`.
@@ -70,7 +127,7 @@
70
127
71
128
##### Admin API
72
129
73
-
Fix issue where consumed worker memory as reported by the `kong.node.get_memory_stats()` PDK method would be incorrectly reported in kilobytes, rather than bytes, leading to inaccurate values in the `/status` Admin API endpoint (and other users of said PDK method).
130
+
-Fix issue where consumed worker memory as reported by the `kong.node.get_memory_stats()` PDK method would be incorrectly reported in kilobytes, rather than bytes, leading to inaccurate values in the `/status` Admin API endpoint (and other users of said PDK method).
74
131
[#6170](https://github.com/Kong/kong/pull/6170)
75
132
76
133
##### Plugins
@@ -88,6 +145,27 @@ Fix issue where consumed worker memory as reported by the `kong.node.get_memory_
88
145
89
146
> Released 2020/07/16
90
147
148
+
:red_circle:**Post-release note (as of 2020/08/13)**: A faulty behavior
149
+
has been observed with this change. When Kong proxies using the balancer
150
+
and a request to one of the upstream `Target` fails, Kong might send the
151
+
same request to another healthy `Target` in a different request later,
152
+
causing response for the failed request to be returned.
153
+
154
+
This bug could be mitigated temporarily by disabling upstream keepalive pools.
155
+
It can be achieved by either:
156
+
157
+
1. In `kong.conf`, set `upstream_keepalive_pool_size=0`, or
158
+
2. Setting the environment `KONG_UPSTREAM_KEEPALIVE_POOL_SIZE=0` when starting
159
+
Kong with the CLI.
160
+
161
+
Then restart/reload the Kong instance.
162
+
163
+
Thanks Nham Le (@nhamlh) for reporting it in [#6212](https://github.com/Kong/kong/issues/6212).
164
+
165
+
:white_check_mark:**Update (2020/08/13)**: A fix to this regression has been
166
+
released as part of [2.1.2](#212). See the section of the Changelog related to this
0 commit comments