Hi Kevin, all,
I tried what you suggested, but AFAICS (and judging from the error
message) supplying these config parameters via the RGW service spec is
not supported right now. Applying it causers an error:
Error EINVAL: Invalid config option request_timeout_ms in spec
The spec looks like this (X.509 cert redacted):
-- 8< --
placement:
count_per_host: 2
label: rgw
service_id: rgw
service_type: rgw
spec:
config:
request_timeout_ms: 120000
rgw_frontend_port: 8000
rgw_frontend_ssl_certificate: '<redacted>'
rgw_realm: myrealm
rgw_zone: myzone
ssl: true
-- 8< --
Having a look at how rgw_frontends is constructed in cephadmservice.py
[1] confirms that this might be unsupported.
Would it make sense to extend the spec so that rgw_frontends can be
configured via cephadm?
Thanks & kind regards,
Thilo
[1] https://github.com/ceph/ceph/blob/57111af8f155e00431f30e0be183eb8f4e6c9eac/src/pybind/mgr/cephadm/services/cephadmservice.py#L881
On Thu, Nov 17, 2022 at 7:27 PM Fox, Kevin M <Kevin.Fox@xxxxxxxx> wrote:
>
> I think you can do it like:
> ```
> service_type: rgw
> service_id: main
> service_name: rgw.main
> placement:
> label: rgwmain
> spec:
> config:
> rgw_keystone_admin_user: swift
> ```
>
> ?
>
> ________________________________________
> From: Thilo-Alexander Ginkel <thilo@xxxxxxxxxx>
> Sent: Thursday, November 17, 2022 10:21 AM
> To: Casey Bodley
> Cc: ceph-users@xxxxxxx
> Subject: Re: Configuring rgw connection timeouts
>
> Check twice before you click! This email originated from outside PNNL.
>
>
> Hello Casey,
>
> On Thu, Nov 17, 2022 at 6:52 PM Casey Bodley <cbodley@xxxxxxxxxx> wrote:
>
> > it doesn't look like cephadm supports extra frontend options during
> > deployment. but these are stored as part of the `rgw_frontends` config
> > option, so you can use a command like 'ceph config set' after
> > deployment to add request_timeout_ms
>
>
> unfortunately that doesn't really seem to work as cephadm is setting the
> config on a service instance level (e.g., client.rgw.rgw.ceph-5.yjgdea), so
> we can't simply override this on a higher hierarchical level. In addition,
> we deploy multiple rgw instances per node (to better utilize available
> resources) which get assigned different HTTP(S) ports by cephadm so they
> can coexist on the same host.
>
> Regards,
> Thilo
> _______________________________________________
> ceph-users mailing list -- ceph-users@xxxxxxx
> To unsubscribe send an email to ceph-users-leave@xxxxxxx
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
