26 May
2020
26 May
'20
2:31 a.m.
Hello everyone
I m trying to implement jaeger tracing in rgw.
I m storing my spans in struct req_state, but some files rgw_sal.cc and
rgw_user.cc dont have anything relation to req_state so I thought of using
extern variable to store req_state, it is wrong because req can run in
parellel, so another way of using it I included data members req_state in
RGWRadosStore and RGWCtrl class then using those to access the req_state
and storing the spans
Is it the right move of adding extra data members in classes?
26 May
26 May
1:41 p.m.
So, this may be more difficult than it first appeared. req_state is
related to requests, and as such, it's only really part of the top-half
of RGW, the OP processing and front-end code. The bottom half, the
storage back-ends (currently only RADOS, but others are coming soon)
doesn't deal directly with the req_state, so it's not passed around.
The dividing line between the top half and the bottom half is rgw_sal.h
and rgw_sal.cc. This is where the back-ends start, and much of the code
in RGW is below that line, and so won't have access to req_state.
It may be that you'll have to pass yous spans into the relevant
functions in the bottom half.
Daniel
On 5/25/20 8:31 PM, Abhinav Singh wrote:
Hello everyone
I m trying to implement jaeger tracing in rgw.
I m storing my spans in struct req_state, but some files rgw_sal.cc and
rgw_user.cc dont have anything relation to req_state so I thought of
using extern variable to store req_state, it is wrong because req can
run in parellel, so another way of using it I included data members
req_state in RGWRadosStore and RGWCtrl class then using those to access
the req_state and storing the spans
Is it the right move of adding extra data members in classes?
_______________________________________________
Dev mailing list -- dev(a)ceph.io
To unsubscribe send an email to dev-leave(a)ceph.io
1:52 p.m.
Perhaps actually passing req_state (I understand why it's not), or
some other more opaque "op context" handle would make sense here?
Matt
On Tue, May 26, 2020 at 7:41 AM Daniel Gryniewicz <dang(a)redhat.com> wrote:
So, this may be more difficult than it first appeared. req_state is
related to requests, and as such, it's only really part of the top-half
of RGW, the OP processing and front-end code. The bottom half, the
storage back-ends (currently only RADOS, but others are coming soon)
doesn't deal directly with the req_state, so it's not passed around.
The dividing line between the top half and the bottom half is rgw_sal.h
and rgw_sal.cc. This is where the back-ends start, and much of the code
in RGW is below that line, and so won't have access to req_state.
It may be that you'll have to pass yous spans into the relevant
functions in the bottom half.
Daniel
On 5/25/20 8:31 PM, Abhinav Singh wrote:
--
Matt Benjamin
Red Hat, Inc.
315 West Huron Street, Suite 140A
Ann Arbor, Michigan 48103
http://www.redhat.com/en/technologies/storage
tel. 734-821-5101
fax. 734-769-8938
cel. 734-216-5309
Hello everyone
I m trying to implement jaeger tracing in rgw.
I m storing my spans in struct req_state, but some files rgw_sal.cc and
rgw_user.cc dont have anything relation to req_state so I thought of
using extern variable to store req_state, it is wrong because req can
run in parellel, so another way of using it I included data members
req_state in RGWRadosStore and RGWCtrl class then using those to access
the req_state and storing the spans
Is it the right move of adding extra data members in classes?
_______________________________________________
Dev mailing list -- dev(a)ceph.io
To unsubscribe send an email to dev-leave(a)ceph.io
_______________________________________________
Dev mailing list -- dev(a)ceph.io
To unsubscribe send an email to dev-leave(a)ceph.io
1428
days inactive
1428
days old
2 comments
3 participants
participants (3)
-
Abhinav Singh -
Daniel Gryniewicz -
Matt Benjamin