[Starlingx-discuss] Ceph upgrading and rebasing
Hi, We are going to upgrade Ceph to 13.2.0. We need to rebase the updated commits in StarlingX on Ceph 13.2.0. We have concerns about the folder wrs. Why do we have this folder? What does it do? 1. Why need to have *.service? Seems Ceph has provided such functions. 2. Ceph.conf: What is this config for? 3. Ceph_pmon_wrapper.sh: where will this script be used? 4. osd-wait-status.py: where does the script be used? And also there are lots of changes related to ceph.spec. Any change explanation/history about these changes? Best wishes Lisa
Hi Lisa, See inline. Ovidiu ________________________________ From: Li, Xiaoyan [xiaoyan.li@intel.com] Sent: Tuesday, September 25, 2018 11:22 AM To: starlingx-discuss@lists.starlingx.io Subject: [Starlingx-discuss] Ceph upgrading and rebasing Hi, We are going to upgrade Ceph to 13.2.0. We need to rebase the updated commits in StarlingX on Ceph 13.2.0. We have concerns about the folder wrs. Why do we have this folder? What does it do? [Ovi] Custom config files needed by stx used by the spec. 1. Why need to have *.service? Seems Ceph has provided such functions. [Ovi] Before systemd, titanium had init scripts which we still use. This was our wrapper to make them work with systemd. Since then ceph added its own, it may be worth to get rid of it and update to the pne in ceph. 2. Ceph.conf: What is this config for? [Ovi] To my knowledge this is the default ceph.conf file used when ceph is enabled. Other config options are appended by puppet to this default (e.g. osd's, mons etc.). 3. Ceph_pmon_wrapper.sh: where will this script be used? [Ovi] This is used by pmon (ceph is pmon mananged) to check the state of ceph services, restart them if they fail or hang. See inside ceph.conf.pmon. The ceph_pmon_wrapper.sh is also related to changes in init-ceph.in (see "WRS: Check if a ceph process is hung" section). 4. osd-wait-status.py: where does the script be used? [Ovi] It's no longer used, it should be removed (we used it in the past to wait for OSDs to shutdown before marking a host as "locked" - now mtc+pmon takes cares of it correctly). References to it should also be removed from init-ceph.in. And also there are lots of changes related to ceph.spec. Any change explanation/history about these changes? [Ovi] Ask Dean Troyer, he did the original push and should have access to the previous commits. Best wishes Lisa
Btw, for what's worth, there is also a ceph.sh in the wrs folder. This file is used by mtc to enable / disable services when a host goes availabel. It is deployed by the rpm in /etc/services.d/ (see spec). When a host is locked/going enabled mtc iterates through all the scripts in /etc/services.d/$host_type where host_type is one in [controller, compute, storage] and executes these scripts with 'start' or 'stop' parameter. This is stx specific, it has nothing to do with systemd. For ceph it also hides its status from pmon view. Regards, Ovidiu ________________________________ From: Poncea, Ovidiu [Ovidiu.Poncea@windriver.com] Sent: Tuesday, September 25, 2018 1:03 PM To: Li, Xiaoyan; starlingx-discuss@lists.starlingx.io Subject: Re: [Starlingx-discuss] Ceph upgrading and rebasing Hi Lisa, See inline. Ovidiu ________________________________ From: Li, Xiaoyan [xiaoyan.li@intel.com] Sent: Tuesday, September 25, 2018 11:22 AM To: starlingx-discuss@lists.starlingx.io Subject: [Starlingx-discuss] Ceph upgrading and rebasing Hi, We are going to upgrade Ceph to 13.2.0. We need to rebase the updated commits in StarlingX on Ceph 13.2.0. We have concerns about the folder wrs. Why do we have this folder? What does it do? [Ovi] Custom config files needed by stx used by the spec. 1. Why need to have *.service? Seems Ceph has provided such functions. [Ovi] Before systemd, titanium had init scripts which we still use. This was our wrapper to make them work with systemd. Since then ceph added its own, it may be worth to get rid of it and update to the pne in ceph. 2. Ceph.conf: What is this config for? [Ovi] To my knowledge this is the default ceph.conf file used when ceph is enabled. Other config options are appended by puppet to this default (e.g. osd's, mons etc.). 3. Ceph_pmon_wrapper.sh: where will this script be used? [Ovi] This is used by pmon (ceph is pmon mananged) to check the state of ceph services, restart them if they fail or hang. See inside ceph.conf.pmon. The ceph_pmon_wrapper.sh is also related to changes in init-ceph.in (see "WRS: Check if a ceph process is hung" section). 4. osd-wait-status.py: where does the script be used? [Ovi] It's no longer used, it should be removed (we used it in the past to wait for OSDs to shutdown before marking a host as "locked" - now mtc+pmon takes cares of it correctly). References to it should also be removed from init-ceph.in. And also there are lots of changes related to ceph.spec. Any change explanation/history about these changes? [Ovi] Ask Dean Troyer, he did the original push and should have access to the previous commits. Best wishes Lisa
participants (2)
-
Li, Xiaoyan
-
Poncea, Ovidiu