If you have similar needs, please leave a message in this issue.

我们公司的项目都是发布在两个机房里面，采用的是异地双机房部署，双写功能，能不能做到原子性呢？

双写针对incr，decr命令怎么保证新加入的redis和原redis数据保持一致？

If you have similar needs, please leave a message in this issue.

我这儿想使用redis的proxy，就分别对camellia-redis-proxy、bilibili-overlord和predixy进行了性能测试对比

配置和这里介绍的一样
https://github.com/netease-im/camellia/blob/master/camellia-redis-proxy/performance-report-1.md

camellia-redis-proxy用的是async模式，用1.0.5版本编译的。可是性能测试结果和您这儿的结果差距有点大。我这儿测试的camellia-redis-proxy对比bilibili-overlord和predixy差了一个数量级。比如单连接时测试get命令，camellia-redis-proxy的平均响应时间是1.5079，而bilibili-overlord和predixy分别是0.17和0.14

不知是不是我用的版本有问题？还是有什么调优的方式？

If you have similar needs, please leave a message in this issue.

If you have similar needs, please leave a message in this issue.

proxy可以k8s部署吗，pod ip变化是否能够感知并自动剔除老pod ip

camellia-redis-proxy如何配置启用ssl来连接到redis cluster？

通过k8s对服务进行部署，报了一个not redis resource的异常，我自己检查代码文件和本地测试都是对的，有点不懂了

请教一下，这个双写，是两个线程同时向两个主发起写，还是按顺序，一个个发起写入？

2021-01-29 03:47:45.763 ERROR 1 --- [ restartedMain] o.s.boot.SpringApplication : Application run failed

org.springframework.beans.factory.BeanCreationException: Error creating bean with name 'commandInvoker' defined in class path resource [com/netease/nim/camellia/redis/proxy/springboot/CamelliaRedisProxyConfiguration.class]: Bean instantiation via factory method failed; nested exception is org.springframework.beans.BeanInstantiationException: Failed to instantiate [com.netease.nim.camellia.redis.proxy.command.CommandInvoker]: Factory method 'commandInvoker' threw exception; nested exception is java.lang.IllegalArgumentException: not found redis-proxy-0.0.1-SNAPSHOT/BOOT-INF/classes!/resource-table.json

If you have similar needs, please leave a message in this issue.

#18

Bug: KeyConverter does not work with the SCAN command

I use CustomKeyConverter implements KeyConverter. It still works with the GET/SET/PUSH/KEY/... commands. But, today, I test with the SCAN command, and it doesn't work the right way. It gets all keys from the Redis resource without the prefix "bid|bgroup" defined

• My CustomKeyConverter class:

public class CustomKeyConverter implements KeyConverter {
  public static final byte[] SEPARATOR = Utils.stringToBytes("|");

  @Override
  public byte[] convert(
      CommandContext commandContext, RedisCommand redisCommand, byte[] originalKey) {
    if (commandContext.getBid() == null || commandContext.getBgroup() == null) {
      return originalKey;
    }
    byte[] bid = Utils.numToBytes(commandContext.getBid(), false);
    byte[] bgroup = Utils.stringToBytes(commandContext.getBgroup());
    byte[] convertedKey =
        new byte
            [bid.length + SEPARATOR.length + bgroup.length + SEPARATOR.length + originalKey.length];
    System.arraycopy(bid, 0, convertedKey, 0, bid.length);
    System.arraycopy(SEPARATOR, 0, convertedKey, bid.length, SEPARATOR.length);
    System.arraycopy(bgroup, 0, convertedKey, bid.length + SEPARATOR.length, bgroup.length);
    System.arraycopy(
        SEPARATOR,
        0,
        convertedKey,
        bid.length + SEPARATOR.length + bgroup.length,
        SEPARATOR.length);
    System.arraycopy(
        originalKey,
        0,
        convertedKey,
        bid.length + SEPARATOR.length + bgroup.length + SEPARATOR.length,
        originalKey.length);
    return convertedKey;
  }

  @Override
  public byte[] reverseConvert(
      CommandContext commandContext, RedisCommand redisCommand, byte[] convertedKey) {
    if (commandContext.getBid() == null || commandContext.getBgroup() == null) {
      return convertedKey;
    }
    byte[] bid = Utils.numToBytes(commandContext.getBid(), false);
    byte[] bgroup = Utils.stringToBytes(commandContext.getBgroup());
    byte[] originalKey =
        new byte
            [convertedKey.length
                - bid.length
                - SEPARATOR.length
                - bgroup.length
                - SEPARATOR.length];
    System.arraycopy(
        convertedKey,
        bid.length + SEPARATOR.length + bgroup.length + SEPARATOR.length,
        originalKey,
        0,
        originalKey.length);
    return originalKey;
  }
}

• Test result:

127.0.0.1:6380> mget k1 k2 k3
1) "tass2"
2) "tass2"
3) "tass3"
127.0.0.1:6380> keys *
1) "k5"
2) "k3"
3) "k4"
4) "k2"
5) "k1"
127.0.0.1:6380> scan 0 match k*         #---> Expect: k1,k2,k3,k4,k5...
1) "60"
2) (empty array)
127.0.0.1:6380> scan 0 count 10         #---> Expect: k1,k2,k3,k4,k5...
1) "60"
2)  1) "ev.monitoring.metric.incrchecked.2212061614"
    2) "a_table|1"
    3) "ev.monitoring.metric.incrchecked.2212061615"
    4) "a_stats_|redis://@127.0.0.1:6379|AsyncCamelliaRedisTemplate|mset|W"
    5) "a_stats_|redis://@127.0.0.1:6379|AsyncCamelliaRedisTemplate|scan|R"
    6) "a_stats_|camellia_t_keys"
    7) "a_stats_|redis://[email protected]:6379|AsyncCamelliaRedisTemplate|command|R"
    8) "a_table|2"
    9) "a_stats_|redis://@127.0.0.1:6379|AsyncCamelliaRedisTemplate|exists|R"
   10) "a_stats_|redis://@127.0.0.1:6379|AsyncCamelliaRedisTemplate|incrby|W"
(0.68s)

时不时的就会报这个错误
ERROR error-log-collector-pool-7-thread-1 c.n.n.c.r.p.u.ErrorLogCollector:62 - com.netease.nim.camellia.redis.proxy.netty.ReplyEncoder:channel not active, remote.ip=/10.120.5.63:47359, count = 1

我记得重启，刚起动完，就会报一下。然后偶尔就会报一个

这个proxy 配制文件是否配制与后端redis 的连接 池。
类似jedis 的 max_active 或者 max_total 这个最大连接数的配制 ？
2、这个到proxy --> redis 连接 ，我看只有3个网络连接。这个是写死的吗?
谢谢

你好，请问下，双写过程如果某个机房网络故障，双写出现问题这种情况怎么补救，有没有办法让数据做到最终一致性，或者该程序是否已经考虑并解决这个问题了？

If you have similar needs, please leave a message in this issue.

Hi,
I do everything like that:

wget https://github.com/netease-im/camellia/releases/download/1.1.7/camellia-redis-proxy-1.1.7.tar.gz
tar zxvf camellia-redis-proxy-1.1.7.tar.gz
cd camellia-redis-proxy-1.1.7/
jar -cvf0M camellia-redis-proxy.jar BOOT-INF/ META-INF/ org/
java -XX:+UseG1GC -XX:+UseContainerSupport -Xms2048m -Xmx2048m -server -jar camellia-redis-proxy.jar --spring.config.location=file:application.yml

When last command run I got this error:

2022-12-02 14:50:03,701 ERROR main  o.s.b.SpringApplication:837 - Application run failed
java.lang.IllegalStateException: Cannot load configuration class: com.netease.nim.camellia.redis.proxy.samples.Application

Please help me.

If you have similar needs, please leave a message in this issue.

最近也在研究redis proxy的方案，你这个项目的About，说了一句和框架没关系的描述，通过关键字redis proxy，都搜索不到，我建议你优化一下项目的About

If you have similar needs, please leave a message in this issue.

日志如下

2022-06-16 14:55:44.958 [INFO] [camellia-schedule-1-8] [camellia.redis.proxy.stats] @@@traceid=N/A@@@ ====spend.stats====
2022-06-16 14:55:44.958 [INFO] [camellia-schedule-1-8] [camellia.redis.proxy.stats] @@@traceid=N/A@@@ command=auth,count=2,avgSpendMs=0.005009,maxSpendMs=0
.005018
2022-06-16 14:55:44.958 [INFO] [camellia-schedule-1-8] [camellia.redis.proxy.stats] @@@traceid=N/A@@@ command=publish,count=2,avgSpendMs=1.170809,maxSpendM
s=1.174107
2022-06-16 14:55:44.958 [INFO] [camellia-schedule-1-8] [camellia.redis.proxy.stats] @@@traceid=N/A@@@ command=get,count=4,avgSpendMs=1.05341025,maxSpendMs=
1.065794
2022-06-16 14:55:44.958 [INFO] [camellia-schedule-1-8] [camellia.redis.proxy.stats] @@@traceid=N/A@@@ command=exists,count=4,avgSpendMs=1.12897225,maxSpend
Ms=1.164166
2022-06-16 14:55:44.958 [INFO] [camellia-schedule-1-8] [camellia.redis.proxy.stats] @@@traceid=N/A@@@ command=brpop,count=3,avgSpendMs=1265.753878,maxSpend
Ms=1598.487612

2022-06-16 14:55:44.958 [INFO] [camellia-schedule-1-8] [camellia.redis.proxy.stats] @@@traceid=N/A@@@ ====upstream.redis.spend.stats====
2022-06-16 14:55:44.958 [INFO] [camellia-schedule-1-8] [camellia.redis.proxy.stats] @@@traceid=N/A@@@ [email protected]:6379,count=15,avgSpendMs=4398.119012933334,maxSpendMs=46122.952246

Welcome to register by company + homepage + use case(optional), your support is important to us.

欢迎采用了Camellia的公司在此登记，您的支持是我们最大的动力。
请按公司名 + 首页 + 应用案例（可选）的格式登记。

camellia用户列表

If you have similar needs, please leave a message in this issue.

如果代理Redis Cluster，Key路由逻辑是什么？

• 和官方一样CRC16，直接把Key路由到正确Slot上
• 和普通客户端一样，随便路由到某个节点，通过 MOVED/ASK 来重定向到正确Slot上

另外，如果代理Redis Cluster，自定义sharding还能用吗？

目前看是用AsyncTaskQueue类支持了pipeline模式。如果一个pipeline中有10个命令，queue中有10个AsyncTask，调用10次callback方法。

如果连续调用10次callback方法，只有第一次生效，其余九次直接返回，如果第一次writeAndFlush时有网络异常导致走到finally里了，但channelInactive又没有触发，是不是会导致pipeline的10条命令一直留在queue中，直到客户端再次请求？而且客户端再次请求时，回复的是上次pipeline中的命令导致异常。如果客户端一段时间内没有再次请求，上一次的pipeline命令会超时失败。

目前没有试过这种情形，想请问下上面这种说法成不成立？

CamelliaIdGenSdkConfig config = new CamelliaIdGenSdkConfig();
config.setUrl("http://127.0.0.1:8081");
config.setMaxRetry(5);//重试次数
CamelliaSnowflakeIdGenSdk idGenSdk = new CamelliaSnowflakeIdGenSdk(config);

建议使用resttemplate等支持eureka的httpclient

您好，现在正在研究云信的redis代理,想咨询一下您以下问题？
1.官方已经推出了Redis 6.0多线程版本，那么云信redis使用netty实现代理的方式，是不是不占优势。
2.官方其实也推出了相关的Redis cluster proxy,那么相比官方的redis cluster proxy云信proxy有什么特别之处？
上面是我的一些问题，只是单纯的想明白云信的优势，然后选择使用。非常感谢您能够解答一下

请问下双读的意思是随机读其中一边的key？还是同时读两边key，取最新的key？
若是同时读两边key，当其中一边删除后，应该如何读取？

dashboard提供的功能过于简单，操作数据，很容易出错？有没有提供监控功能的计划呢？以后对于这个项目还有哪些功能的升级？

2021-02-24 05:41:27.934 ERROR 1 --- [pool-6-thread-1] c.n.n.c.r.proxy.util.ErrorLogCollector : com.netease.nim.camellia.redis.proxy.netty.ReplyEncoder:channel not active, remote.ip=/127.0.0.1:51147, count = 30

2021-02-24 05:41:27.934 ERROR 1 --- [pool-6-thread-1] c.n.n.c.r.proxy.util.ErrorLogCollector : com.netease.nim.camellia.redis.proxy.netty.ReplyEncoder:channel not active, remote.ip=/127.0.0.1:49193, count = 30

2021-02-24 05:41:27.934 ERROR 1 --- [pool-6-thread-1] c.n.n.c.r.proxy.util.ErrorLogCollector : com.netease.nim.camellia.redis.proxy.netty.ReplyEncoder:channel not active, remote.ip=/127.0.0.1:51941, count = 30

2021-02-24 05:41:27.934 ERROR 1 --- [pool-6-thread-1] c.n.n.c.r.proxy.util.ErrorLogCollector : com.netease.nim.camellia.redis.proxy.netty.ReplyEncoder:channel not active, remote.ip=/127.0.0.1:52119, count = 30

2021-02-24 05:41:27.934 ERROR 1 --- [pool-6-thread-1] c.n.n.c.r.proxy.util.ErrorLogCollector : com.netease.nim.camellia.redis.proxy.netty.ReplyEncoder:channel not active, remote.ip=/127.0.0.1:49754, count = 30

2021-02-24 05:41:27.934 ERROR 1 --- [pool-6-thread-1] c.n.n.c.r.proxy.util.ErrorLogCollector : com.netease.nim.camellia.redis.proxy.netty.ReplyEncoder:channel not active, remote.ip=/127.0.0.1:49872, count = 30

2021-02-24 05:41:27.934 ERROR 1 --- [pool-6-thread-1] c.n.n.c.r.proxy.util.ErrorLogCollector : com.netease.nim.camellia.redis.proxy.netty.ReplyEncoder:channel not active, remote.ip=/127.0.0.1:51707, count = 30

com.netease.nim.camellia.redis.proxy.command.async.AsyncCamelliaSimpleClient#preheat
log some info, without any log level check

Camellia is using dependencies:

Maven: org.apache.logging.log4j:log4j-api:2.10.0
Maven: org.apache.logging.log4j:log4j-to-slf4j:2.10.0

Please recheck and upgrade the version of log4j:

• https://logging.apache.org/log4j/2.x/security.html
• https://geekflare.com/scan-and-fix-log4j-vulnerability/

Update Error Code API /permission/ip-checkers
[x] Not Found Ip Checker configuration -> Error code: 404
[x] Check existed by bid and bgroup before updating the Ip-Checker (prevent update bid and bgroup)

My fault, I'm going to update those issues tonight or tomorrow.

Hello
Could someone explain why there is no support for selecting other database than 0?

Or maybe it is possible to get some workaround by setup database in "resource-table.json" file and launch few instances of redis-proxy (db0 on port 6000, db1 on port 6001 etc)?
For example, something like that:

server:
  port: 6002
spring:
  application:
    name: camellia-redis-proxy-server

camellia-redis-proxy:
  database: 2 # <---- here setup database id that is supported, default=0
  transpond:
    type: local
    local:
      type: complex
      json-file: /tmp/resource-table.json

and:

{
  "type": "simple",
  "operation": {
    "read": "redis://[email protected]:6379/2", # <--- here is "destination" database id
    "type": "rw_separate",
    "write": {
      "resources": [
        "redis://[email protected]:6379/2", # <--- here is "destination" database id
        "redis://[email protected]:6380/2", # <--- here is "destination" database id
      ],
      "type": "multi"
    }
  }
}

Thanks for help

Hi @caojiajun, I want to say camellia is an excellent project. Thank you all. <3

Follow by performance-report.md, I try to test performance by redis-benchmark, but I got an issue:

 ~ redis-benchmark -n 1000000 -r 10000 -c 200 -t set,get,incr,lpush,rpush,lpop,rpop,sadd,hset,spop,lpush,lrange -q -p 6380 -h [myhost] -a [pass]
ERROR: Not Support
ERROR: failed to fetch CONFIG from [myhost]:6380
WARNING: Could not fetch server CONFIG
SET: rps=0.0 (overall: 0.0) avg_msec=nan (overall: nan) 

and I read the code of this project do not support the CONFIG redis command:
CONFIG(CommandSupportType.NOT_SUPPORT, null, null, Blocking.FALSE, null),

How can you benchmark camellia-redis-proxy with redis-benchmark.

Again, Thank you so much!

why does this method always return null if withCRKF input is a false value

        long originalNum = 10;
        byte[] bytes = Utils.numToBytes(originalNum, false); // --> bytes = null ?

I try to upgrade camellia to version v1.1.3, so I found an error Maven dependency pulling a wrong dependency (maybe since v.1.1.2, I test with v1.1.1 still works)

....
		<dependency>
			<groupId>com.netease.nim</groupId>
			<artifactId>camellia-redis-proxy-spring-boot-starter</artifactId>
			<version>1.1.3</version>
		</dependency>

....

• V1.1.1:
  

• V1.1.3:
  

==> The consequence is cannot load classes of camellia-redis-proxy module

你好，

我看到camellia redis proxy支持设定密码，我想了解对应redis cluster proxy里的auth口令部分，在camellia redis proxy中应该如何设置？

如题

我后端对接的是pika 。报这个错误 有点不理解 。看连接 ，还连上来了，最后总是报这么个错误 。。。
ERROR error-log-collector-pool-7-thread-1 c.n.n.c.r.p.u.ErrorLogCollector:62 - com.netease.nim.camellia.redis.proxy.command.async.RedisClient:RedisClient[@10.10.251.11:6379][id=53] stopping, command return NOT_AVAILABLE, count = 1

1. 背景

redis proxy 作为一个大型redis cluster集群的proxy。
该cluster 为110个master 110个slave。
proxy节点数为10个节点。在客户端使用lettuce作为连接，并根据请求进行了负载均衡。
压测VUSER 1200

2. 现象

压测到一定流量后，出现no child group等错误。该问题应该是netty的child group无法再处理了。

3. 问题分析

发现在源码中 soBacklog的值设置为1024.该值在大集群海量并发情况下，是否合理？不确定该问题是否由这个设置导致的。

4. 期望

是否可以提供一些针对不同场景的参数最佳实践？

启动运行一会之后，打这个日志，这个问题产生的原因是什么

We should immediately renew slots when having any node down, but currently it is not

你好，想请教下，目前不支持的命令比如DUMP这种，不支持的原因是什么呀？比如RESTORE命令？
我这边想用camellia做一对多数据同步的写通道，但是做全量同步的时候需要用到RESTORE命令。

Enhancement: Support connect limit for each tenant(bid, bgroup) by only 1 global configuration

We currently have the connect limit configuration for:

• all tenants: max.client.connect=10000
• each tenant by bid and bgroup: 1.default.max.client.connect=10000

Maybe we need to support config limit connection for each tenant by only 1 "global configuration":

• Global tenant: bid=-1 & bgroup=default
• -1.default.max.client.connect=10000 ==> this configuration will apply for all tenants if those are not set.

I will commit to this update today.

hi @caojiajun,
Now we're having ipCheckerPlugin - A plugin for IP black and white list restrictions on clients accessing the proxy. But It still read the configuration from file camellia-redis-proxy.properties.
Should we provide a new Ip Checker Plugin: dynamicIpCheckerPlugin, It will read configurations via camellia-dashboard like how we fetch data from that. We will store these configurations in the database instead of a file.

create table camellia_ip_checker
(
    id          bigint(64) auto_increment comment 'Auto increment field'
        primary key,
    bid         bigint(64)    null comment 'bid',
    bgroup      varchar(64)   null comment 'bgroup',
    mode        tinyint(1)    not null comment '0=UNKNOWN, 1=BLACK, 2=WHITE',
    ip_list     varchar(1024) not null comment 'support ip, also supports network segment, comma separated.ex:2.2.2.2,5.5.5.5,3.3.3.0/24,6.6.0.0/16',
    create_time varchar(64)   null comment 'create time',
    update_time varchar(64)   null comment 'Update time',
    constraint bid_bgroup_unique
        unique (bid, bgroup)
)
    comment 'IP checker table' charset = utf8;

CREATE TABLE `camellia_resource_info` (
  `id` bigint(64) NOT NULL primary key auto_increment comment '自增字段',
  `url` varchar(1024) NOT NULL comment '资源url',
  `info` varchar(1024) NOT NULL comment '描述',
  `tids` varchar(1024) DEFAULT NULL comment '引用的tids',
  `create_time` varchar(2000) DEFAULT NULL comment '创建时间',
  `update_time` varchar(64) DEFAULT NULL comment '更新时间'
) ENGINE=InnoDB DEFAULT CHARSET=utf8 COMMENT='资源信息表';

As the title, why are data types 'create_timeand update_time` varchar?

Camellia-dashboard: I implement api-key authentication. Every request must have the header api-key: xxx for more security.

Camellia-redis-proxy: I don't want to modify your code, how can I override Request Interceptor of CamelliaApi (2 methods: getResourceTable() , reportStats())

I saw your code init Camellia Feign Client inside CamelliaApiUtil:

public class CamelliaApiUtil {

    public static CamelliaApi init(String url, int connectTimeoutMillis, int readTimeoutMillis) {
        return Feign.builder()
                .encoder(new JacksonEncoder())
                .decoder(new JacksonDecoder())
                .requestInterceptor(template -> {
                    if (CamelliaApiEnv.source != null) {
                        template.header(CamelliaApiEnv.REQUEST_SOURCE, CamelliaApiEnv.source);
                    }
                })
                .options(new Request.Options(connectTimeoutMillis, readTimeoutMillis))
                .target(CamelliaApi.class, url);
    }
}

But it is not Bean, so I cannot override it or add Request Interceptor like that:

    @Bean
    public RequestInterceptor myCustomInterceptor() {
        return template->template.header("api-key", "xxx");
    }

I suggest a feature that we can config headers inside transpond RemoteProperties:

application.yaml

camellia-redis-proxy:
  transpond:
    type: remote
    remote: 
      url: ${CAMELLIA_DASHBOARD_URL:http://127.0.0.1:8080} #camellia-dashboard's address
      check-interval-millis: ${REMOTE_CHECK_INTERVAL_MILLIS:5000} # Polling period to camellia-dashboard
      dynamic: ${REMOTE_DYNAMIC:true} # indicates that multiple sets of configurations are supported, the default is true
      api-key-header-name: ${REMOTE_API_KEY_HEADER_NAME:api-key} # API Key header name
      api-key: ${REMOTE_API_KEY:xxx} # API Key

And add it inside :

public class CamelliaApiUtil {

    public static CamelliaApi init(String url, String apiKeyHeaderName, String apiKey, int connectTimeoutMillis, int readTimeoutMillis) {
        return Feign.builder()
                .encoder(new JacksonEncoder())
                .decoder(new JacksonDecoder())
                .requestInterceptor(template -> {
                    if (CamelliaApiEnv.source != null) {
                        template.header(CamelliaApiEnv.REQUEST_SOURCE, CamelliaApiEnv.source);
                    }
                    if(apiKeyHeaderName != null){ // <-------------------------------------------this
                        template.header(apiKeyHeaderName, apiKey);
                    }
                })
                .options(new Request.Options(connectTimeoutMillis, readTimeoutMillis))
                .target(CamelliaApi.class, url);
    }
}