Thread: [GENERAL] RemoveIPC problem
Thanks to a typo, I did not turn off systemd's RemoveIPC, and had many many pg restarts before I figured out the problem. Should my data be OK? Or do I need to dump & reload? -- Scott Ribe scott_ribe@elevated-dev.com (303) 722-0567
scott ribe <scott_ribe@elevated-dev.com> writes: > Thanks to a typo, I did not turn off systemd's RemoveIPC, and had many many pg restarts before I figured out the problem. > Should my data be OK? Or do I need to dump & reload? I don't know of any reason to think that that poses a data corruption risk. (But I've been wrong before.) regards, tom lane
Yeah, I was kind of thinking that PG detects the semaphore not existing, bails immediately, restarts clean, thus no problem.I just wanted to hear from people, like you, that know way more than I do about the internals. > On Aug 31, 2017, at 9:08 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote: > > scott ribe <scott_ribe@elevated-dev.com> writes: >> Thanks to a typo, I did not turn off systemd's RemoveIPC, and had many many pg restarts before I figured out the problem. > >> Should my data be OK? Or do I need to dump & reload? > > I don't know of any reason to think that that poses a data corruption > risk. (But I've been wrong before.) > > regards, tom lane -- Scott Ribe scott_ribe@elevated-dev.com (303) 722-0567
On Fri, Sep 1, 2017 at 12:10 PM, scott ribe <scott_ribe@elevated-dev.com> wrote: > Yeah, I was kind of thinking that PG detects the semaphore not existing, bails immediately, restarts clean, thus no problem.I just wanted to hear from people, like you, that know way more than I do about the internals. As long as you don't run Postgres on scissors with things like fsync = off or full_page_writes = off, there should be no risk with the data consistency. -- Michael
Yeah, not my style. fsync is on, no caching RAID controller, etc. Thanks. > On Aug 31, 2017, at 9:48 PM, Michael Paquier <michael.paquier@gmail.com> wrote: > > As long as you don't run Postgres on scissors with things like fsync = > off or full_page_writes = off, there should be no risk with the data > consistency. -- Scott Ribe scott_ribe@elevated-dev.com (303) 722-0567