| Bug #93004 | MGR failed to boot in VPC(ECS) when using public ip | ||
|---|---|---|---|
| Submitted: | 30 Oct 2018 7:24 | Modified: | 2 Jul 2019 1:09 |
| Reporter: | Zhenghu Wen (OCA) | Email Updates: | |
| Status: | Can't repeat | Impact on me: | |
| Category: | MySQL Server: Group Replication | Severity: | S4 (Feature request) |
| Version: | 5.7.20 | OS: | Linux |
| Assigned to: | MySQL Verification Team | CPU Architecture: | Any |
[5 Nov 2018 1:10]
Zhenghu Wen
Hi Bogdan： if it could be verified in your environment？
[5 Nov 2018 1:43]
MySQL Verification Team
Hi, I do not have access to aliyun and on a normal setup I can't reproduce this. Now if you look at your ip addr output there is no public ip there so I'm not sure if I can call this a bug.. If you are running the GR inside a cloud where your boxes have private ip's and are seen trough 1:1 nat from the world I believe they would have to be configured to use the private ip's not public ones as they don't see public ones properly. all best Bogdan
[5 Nov 2018 3:33]
Zhenghu Wen
may be it could be a feature request
[5 Nov 2018 14:56]
MySQL Verification Team
Hi, I'm not sure FR makes sense either. You have a system where N mysql servers have private IP's and they see each other with/trough those private IP's. Then you have those private IP's 1:1 natted to the real world. You want to configure your Innodb cluster to use those private ip's and that works. I don't see where does the real IP plays a role here since it's not visible by any of the servers. all best Bogdan
[2 Jul 2019 1:09]
Zhenghu Wen
same as https://bugs.mysql.com/bug.php?id=92665
[9 Feb 2023 14:33]
MySQL Verification Team
Solution from Bug#109996 should fix this bug too
Description: MGR xcom will: 1. check the group_replication_local_address parameter in bool is_parameters_syntax_correct(const Gcs_interface_parameters &interface_params) by using get_ipv4_local_addresses(std::map<std::string, int>& addr_to_cidr_bits, bool filter_out_inactive) 2. set site_def->nodeno in void site_install_action(site_def *site, cargo_type operation) by using node_no xcom_find_node_index(node_list *nodes) 3. check if a server is itself in static server * mksrv(char *srv, xcom_port port) by using node_no xcom_mynode_match(char *name, xcom_port port) all of three will get server ipaddress by using static int init_sock_probe(sock_probe *s) but in cloud, a vpc ecs init_sock_probe() can only get a loop ip and a private ip, it could not get the public ip . so if we using the public ip when setup multi mgr node cross vpc. it will failed. How to repeat: 1. create a vpc ecs in aliyun or others with public ip (it is default option). mg-node0 公网 IP：59.111.148.50私有网 IP：192.168.129.143 root@mg-node0:~# ip a 1: lo: <LOOPBACK,UP,LOWER_UP> mtu 16436 qdisc noqueue state UNKNOWN link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00 inet 127.0.0.1/8 scope host lo inet6 ::1/128 scope host valid_lft forever preferred_lft forever 2: eth0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc pfifo_fast state UP qlen 1000 link/ether fa:16:3e:2e:50:70 brd ff:ff:ff:ff:ff:ff inet 192.168.129.143/17 brd 192.168.255.255 scope global eth0 inet6 fe80::f816:3eff:fe2e:5070/64 scope link valid_lft forever preferred_lft forever but it indeed has a public ip address. 2. boot mgr by seting group_replication_local_address as the public ip loose-group_replication_local_address= "59.111.148.50:3307" when start group_replication, it failed, log likes: [ERROR] Plugin group_replication reported: '[GCS] There is no local IP address matching the one configured for the local node (59.111.148.50:3307).'