To add to this I’ve noticed that for RGW with s3website api enabled it’s even worse - sometimes it needs a hard reset. Tested on 17.2.6. Ondra

On 26. 8. 2024, at 4:48, Huy Nguyen <viplanghe6@gmail.com> wrote:

Hi community, I'm using ceph v18.2.4. Each time I commit a period, all of my radosgw instances paused for about 30s which is kinda long, and make my object service down during that time.

how to avoid the down time? Is there better/safer practice?

Thanks

Here is radosgw logs each time I did a period commit. You can see it was paused for a while since the last request

2024-08-23T22:43:32.152+0700 7f4594d2f700 1 ====== starting new request req=0x7f42dcfbf730 ===== 2024-08-23T22:43:32.152+0700 7f438b11c700 1 ====== req done req=0x7f42dcfbf730 op status=0 http_status=200 latency=0.000000000s ====== 2024-08-23T22:43:32.152+0700 7f438b11c700 1 beast: 0x7f42dcfbf730: 172.24.9.101 - anonymous [23/Aug/2024:22:43:32.152 +0700] "GET / HTTP/1.0" 200 231 - "check_http/v2.2 (monitoring-plugins 2.2)" - latency=0.000000000s 2024-08-23T22:44:02.778+0700 7f42e1fca700 1 rgw realm reloader: Pausing frontends for realm update... 2024-08-23T22:44:02.778+0700 7f42e1fca700 1 rgw realm reloader: Frontends paused 2024-08-23T22:44:03.358+0700 7f42e1fca700 1 rgw realm reloader: driver closed 2024-08-23T22:44:03.497+0700 7f42e1fca700 1 rgw realm reloader: Creating new driver 2024-08-23T22:44:03.497+0700 7f42e1fca700 1 mgrc service_daemon_register rgw.5055090799 metadata {arch=x86_64,ceph_release=reef,ceph_version=ceph version 18.2.4 (e7ad5345525c7aa95470c26863873b581076945d) reef (stable),ceph_version_short=18.2.4.... 2024-08-23T22:44:03.497+0700 7f42e1fca700 1 rgw realm reloader: Finishing initialization of new driver 2024-08-23T22:44:03.497+0700 7f42e1fca700 1 rgw realm reloader: - REST subsystem init 2024-08-23T22:44:03.497+0700 7f42e1fca700 1 rgw realm reloader: - usage subsystem init 2024-08-23T22:44:03.497+0700 7f42e1fca700 1 rgw realm reloader: Resuming frontends with new realm configuration. 2024-08-23T22:44:06.135+0700 7f4637674700 1 ====== starting new request req=0x7f42dcfbf730 ===== 2024-08-23T22:44:06.136+0700 7f4637674700 -1 res_query() failed 2024-08-23T22:44:06.136+0700 7f4637674700 1 ====== req done req=0x7f42dcfbf730 op status=0 http_status=405 latency=0.000999984s ====== 2024-08-23T22:44:06.136+0700 7f4637674700 1 beast: 0x7f42dcfbf730: 172.24.17.21 - - [23/Aug/2024:22:44:06.135 +0700] "GET /admin HTTP/1.1" 405 133 - "curl/7.61.1" - latency=0.000999984s _______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io