View Issue Details
| ID | Project | Category | View Status | Date Submitted | Last Update |
|---|---|---|---|---|---|
| 0000665 | Pgpool-II | Bug | public | 2020-12-01 10:27 | 2020-12-02 15:09 |
| Reporter | omsupport | Assigned To | pengbo | ||
| Priority | normal | Severity | minor | Reproducibility | always |
| Status | feedback | Resolution | open | ||
| Product Version | 4.1.5 | ||||
| Summary | 0000665: pcp_attach_node is looking fro wrong directory | ||||
| Description | Every time I try to reacttach a failed node with this command, pcp_attach_node -n 1 -U pgpool -w I'm now getting ERROR: connection to socket "/var/run/postgresql/.s.PGSQL.9898" failed with error "No such file or directory" This is NOT how it used to work. I have it set to the /tmp directory as default in pgpool.conf. Problem is I have never changed this. I have automation setup for over a year to reattach node if the actual backend is healthy but then it stopped working after the last few version updates. The comments say it defaults to /var/run/postgresql for Debian. I'm running Centos 7 so it should be using /tmp as it is declared. It always has. The actual file does show in the /tmp directory but it's looking for it in the wrong place. I shouldn't have to specify the path in the command like "pcp_attach_node -h /tmp -n 1 -U pgpool -w" when I have it set to the default pcp_socket_dir parameter. I believe something broken. pcp_socket_dir = '/tmp' # Unix domain socket path for pcp # The Debian package defaults to # /var/run/postgresql # (change requires restart) | ||||
| Steps To Reproduce | try to reattach node in Centos 7 without specifying the -h parameter. | ||||
| Tags | No tags attached. | ||||
|
|
It is the normal behavior. Since 4.1.3, UNIX_DOMAIN_PATH is changed to "/var/run/postgresql". pcp command will check /var/run/postgresql by default. If you have specified pcp_socket_dir = '/tmp', you need to specify the path using "-h" option explicitly. [Release note] - Change PCP UNIX_DOMAIN_PATH of RPM package to /var/run/postgresql. https://www.pgpool.net/docs/latest/en/html/release-4-1-3.html pcp_socket_dir = '/tmp' is the configuration for pgpool server side only and it does not affect the pcp command. |
|
|
Well the problem is the pgpool.conf has /tmp by default. As I said I never changed the location. The location of the .s.PGSQL.9898 file still shows in /tmp. So why would we think to look in /var/run/postgresql or use ? How was "Change PCP UNIX_DOMAIN_PATH of RPM package to /var/run/postgresql. (Bo Peng)" a bug? That seems like more of a change. Does the pgpool.conf file have updated path on new installs? |
|
|
> Well the problem is the pgpool.conf has /tmp by default. As I said I never changed the location. The location of the .s.PGSQL.9898 file still shows in /tmp. > So why would we think to look in /var/run/postgresql or use ? How was "Change PCP UNIX_DOMAIN_PATH of RPM package to /var/run/postgresql. (Bo Peng)" a bug? > That seems like more of a change. Does the pgpool.conf file have updated path on new installs? Because the pgpool.conf installed from RPM package has the default value pcp_socket_dir = '/var/run/postgresql', we changed PCP UNIX_DOMAIN_PATH in RPM packages only. pcp_socket_dir = '/var/run/postgresql' If you are installing from source code, pgpool will not be affected by this change. |
| Date Modified | Username | Field | Change |
|---|---|---|---|
| 2020-12-01 10:27 | omsupport | New Issue | |
| 2020-12-01 12:42 | pengbo | Note Added: 0003621 | |
| 2020-12-01 12:42 | pengbo | Assigned To | => pengbo |
| 2020-12-01 12:42 | pengbo | Status | new => feedback |
| 2020-12-01 12:56 | omsupport | Note Added: 0003622 | |
| 2020-12-01 12:56 | omsupport | Status | feedback => assigned |
| 2020-12-02 15:08 | pengbo | Note Added: 0003631 | |
| 2020-12-02 15:09 | pengbo | Status | assigned => feedback |
