Hi
I’m not sure why AMFND isn’t an applier. Perhaps because the AMF implementation
was done before IMM. Praveen/Nagu?
Thanks
Gary
-----Original Message-----
From: A V Mahesh <mahesh.va...@oracle.com>
Organization: Oracle Corporation
Date: Thursday, 22 June 2017 at 6:58 pm
To: <anders.bjornerst...@telia.com>, gary <gary....@dektech.com.au>
Cc: <opensaf-devel@lists.sourceforge.net>, <minh.c...@dektech.com.au>
Subject: Re: [devel] [PATCH 1/1] amfnd: retry on ERR_NOT_EXIST [#2490]
Hi Gary,
I am also in sync with Anders, based on my previous understanding of IMM
implementation (when I was IMM maintainer )
we have a concept called multiple applier, so multiple IMMA can register
for a class , so it will easy for AMFND to get notification by
registering as applier.
In general, it is application (AMFND) responsibility to make sure that
object existence before doing Admin operation as if you are triggering a
requests two different amfnd node.
For now you can take help from Zoran/Hung, they can provide you quick
solution , in future (once I sync up with latest IMM code) you can
expect solution form me as well :)
-AVM
On 6/22/2017 12:18 PM, and...@acm.org wrote:
> Hi,
>
> The imm "can guarantee" that a CCB than has been ack'ed as commited to
the OM ccb-client (or an OI-applier), has been committed in fevs order over
all nodes.
> In particular it means that it has been commited at the node where the
reply was provided.
> If you send messages between nodes based on that commit, then to ensure
that a receiver at a remote node receives that mesage in fevs
> order, then you need to use a fevs ordered messaging service, e.g.
admin-operation.
> Note an admin-operaton request goes over fevs, but the reply does not.
> This means that the receiver of the request will receive the request in
fevs order.
> But the receiver of the reply is only guaranteed that a reply is
returned after the request has ben procesd in fevs order.
>
> /AndersBj
>
>
>> ----Ursprungligt meddelande----
>> Från : gary....@dektech.com.au
>> Datum : 2017-06-22 - 07:24 (WEST)
>> Till : mahesh.va...@oracle.com
>> Kopia : minh.c...@dektech.com.au, opensaf-devel@lists.sourceforge.net
>> Ämne : Re: [devel] [PATCH 1/1] amfnd: retry on ERR_NOT_EXIST [#2490]
>>
>> Hi Mahesh
>>
>> I don’t know of another way to fixing this issue other than what is
already proposed in this ticket, unless IMM can guarantee that a commit has
been applied to all nodes.
>>
>> As discussed, the object create and admin op are done on the same node.
But that may trigger a request to amfnd on another node, where that object may
not yet exist.
>>
>> Thanks
>> Gary
>>
>> -----Original Message-----
>> From: A V Mahesh <mahesh.va...@oracle.com>
>> Organization: Oracle Corporation
>> Date: Thursday, 22 June 2017 at 12:32 pm
>> To: gary <gary....@dektech.com.au>
>> Cc: praveen malviya <praveen.malv...@oracle.com>,
<hans.nordeb...@ericsson.com>, Nagendra Kumar <nagendr...@oracle.com>, minh
chau <minh.c...@dektech.com.au>, <opensaf-devel@lists.sourceforge.net>
>> Subject: Re: [devel] [PATCH 1/1] amfnd: retry on ERR_NOT_EXIST [#2490]
>>
>> Hi Gary,
>>
>>
>> On 6/21/2017 2:32 PM, Gary Lee wrote:
>> > Hi
>> >
>> > I should have written that amfd acts on the admin op by sending a
msg to a remote amfnd.
>> Sorry ,I didn't get that either you asked me question or you
informed
>> me alternatively how you are planning to fix,
>> if you asked me a question can you please elaborate.
>>
>> -AVM
>>
>> >
>> > Gary
>> >
>> >> On 21 Jun 2017, at 6:55 pm, A V Mahesh <mahesh.va...@oracle.com>
wrote:
>> >>
>> >> Hi Gary,
>> >>
>> >>> On 6/21/2017 12:47 PM, Gary Lee wrote:
>> >>> Hi Mahesh
>> >>>
>> >>> The IMM objects and admin op are issued on the same node.
>> >>> AMFD will ‘forward’ the admin op to AMFND on *another* node,
where sometimes the object is not yet available.
>> >> This explains , their is a possibility AMFND on *another* node
can initiated admin op earlier than the IMM object sync
>> >> why because AMF messages are having higher priority (MDS).
>> >>
>> >> Can you please explore option of both object create & admin
operation can occur on same AMFND node ( not ‘forwarding’ )
>> >>
>> >> -AVM
>> >>> Thanks
>> >>> Gary
>> >>>
>> >>> On 21/6/17, 5:14 pm, "A V Mahesh" <mahesh.va...@oracle.com>
wrote:
>> >>>
>> >>> Hi Pravee/Gary,
>> >>> On 6/21/2017 12:13 PM, praveen malviya wrote:
>> >>> > Hi Gary,
>> >>> >
>> >>> > Is there any ticket in IMM for any related issue?
>> >>> > Both creation of SU and admin operation on it will go
through IMM only.
>> >>> > I do not know how IMM handles it and how IMM spec talks
about it.
>> >>> [AVM] Even we have problem of newly created IMM object
is taking some
>> >>> time to be available on other nodes
>> >>> it will be a implementation problem and it is
not related
>> >>> to IMM specification .
>> >>> > Does IMM spec allow admin operation on a entity for
which all synced
>> >>> > IMMNDS (leaving non synced IMMNDS on the nodes joining
the cluster)
>> >>> > are not updated?
>> >>> [AVM] Again, this is also not related to IMM
specification , if we
>> >>> resolve above this issue will not rise .
>> >>> Gary, Is this issue you are facing in case where you
are creating object
>> >>> on one node and admin operation performed on different
node ?
>> >>> or on same same node ?
>> >>> -AVM
>> >>> On 6/21/2017 12:30 PM, Gary Lee wrote:
>> >>> > Hi Praveen
>> >>> >
>> >>> > From talking with Zoran/Hung, the IMM behaviour is
expected and correct.
>> >>> > I’m not sure about the last question.
>> >>> >
>> >>> > As you know, in AMFND, the trigger to read from IMM is
from an AMFD message.
>> >>> > The CCB could be committed on the node where active AMFD
resides, so I guess the admin op would be accepted regardless of the answer to
your question.
>> >>> >
>> >>> > Gary
>> >>> >
>> >>> > On 21/6/17, 4:43 pm, "praveen malviya"
<praveen.malv...@oracle.com> wrote:
>> >>> >
>> >>> > Hi Gary,
>> >>> >
>> >>> > Is there any ticket in IMM for any related issue?
>> >>> > Both creation of SU and admin operation on it will
go through IMM only.
>> >>> > I do not know how IMM handles it and how IMM spec
talks about it.
>> >>> > Does IMM spec allow admin operation on a entity for
which all synced
>> >>> > IMMNDS (leaving non synced IMMNDS on the nodes
joining the cluster) are
>> >>> > not updated?
>> >>> >
>> >>> > Thanks,
>> >>> > Praveen
>> >>> >
>> >>> > On 15-Jun-17 1:11 PM, Gary Lee wrote:
>> >>> > > On a congested network, sometimes a newly created
IMM object can take some
>> >>> > > time to be available on other nodes.
>> >>> > >
>> >>> > > In our test, a new SU is created on SC-1 and
unlocked. But sometimes
>> >>> > > it fails on a remote node due to:
>> >>> > >
>> >>> > > 2017-05-19 13:55:19 SC-2 osafamfnd[258]: ER
amf_saImmOmSearchInitialize_o2 failed: 12
>> >>> > >
>> >>> > > To get around this, we will retry on
SA_AIS_ERR_NOT_EXIST a few times.
>> >>> > > ---
>> >>> > > src/amf/amfnd/util.cc | 15 +++++++++++++++
>> >>> > > 1 file changed, 15 insertions(+)
>> >>> > >
>> >>> > > diff --git a/src/amf/amfnd/util.cc
b/src/amf/amfnd/util.cc
>> >>> > > index ed0905ce2..bca642eac 100644
>> >>> > > --- a/src/amf/amfnd/util.cc
>> >>> > > +++ b/src/amf/amfnd/util.cc
>> >>> > > @@ -38,6 +38,9 @@
>> >>> > > #include <unistd.h>
>> >>> > > #include "osaf/configmake.h"
>> >>> > > #include "amf/amfnd/avnd.h"
>> >>> > > +#include "base/osaf_time.h"
>> >>> > > +
>> >>> > > +extern struct ImmutilWrapperProfile
immutilWrapperProfile;
>> >>> > >
>> >>> > > const char *presence_state[] = {
>> >>> > > "OUT_OF_RANGE", "UNINSTANTIATED",
"INSTANTIATING",
>> >>> > > @@ -335,6 +338,18 @@ SaAisErrorT
amf_saImmOmSearchInitialize_o2(
>> >>> > >
scope, searchOptions, searchParam,
>> >>> > >
attributeNames, &searchHandle);
>> >>> > > }
>> >>> > > + } else if (rc == SA_AIS_ERR_NOT_EXIST) {
>> >>> > > + // it is possible for 'rootName' to be not
yet available
>> >>> > > + // at the local immnd. Retry a few times to
allow CCB to be propagated.
>> >>> > > + unsigned int nTries = 1;
>> >>> > > + while (rc == SA_AIS_ERR_NOT_EXIST &&
>> >>> > > + nTries < immutilWrapperProfile.nTries) {
>> >>> > > + osaf_nanosleep(&kHundredMilliseconds);
>> >>> > > + rc =
immutil_saImmOmSearchInitialize_o2(immHandle, rootName.c_str(),
>> >>> > > + scope, searchOptions, searchParam,
>> >>> > > + attributeNames, &searchHandle);
>> >>> > > + nTries++;
>> >>> > > + }
>> >>> > > }
>> >>> > > return rc;
>> >>> > > }
>> >>> > >
>> >>> >
>> >>> >
>> >>> >
>> >>> >
>> >>> >
------------------------------------------------------------------------------
>> >>> > Check out the vibrant tech community on one of the
world's most
>> >>> > engaging tech sites, Slashdot.org!
https://urldefense.proofpoint.com/v2/url?u=http-3A__sdm.link_slashdot&d=DwIGaQ&c=RoP1YumCXCgaWHvlZYR8PQcxBKCX5YTpkKY057SbK10&r=Ua8D9LRbRt6_3fFyQghU61tppZpWmDKj9QGctDqRYQQ&m=3lVa-ymN89sIVfvS57TSd0edM74_gBo3GkkdaSxuTCA&s=WL9nnf9tO-B21XeTvgltME0cARW6eq0IpDLuJQocH_k&e=
>> >>> > _______________________________________________
>> >>> > Opensaf-devel mailing list
>> >>> > Opensaf-devel@lists.sourceforge.net
>> >>> >
https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.sourceforge.net_lists_listinfo_opensaf-2Ddevel&d=DwIGaQ&c=RoP1YumCXCgaWHvlZYR8PQcxBKCX5YTpkKY057SbK10&r=Ua8D9LRbRt6_3fFyQghU61tppZpWmDKj9QGctDqRYQQ&m=3lVa-ymN89sIVfvS57TSd0edM74_gBo3GkkdaSxuTCA&s=5AyrQ2cwgH0HW6VfyB4nxYnHjDz-L_xk5n3HjZmw350&e=
>> >>>
>> >>>
>>
>>
>>
>>
>>
>>
------------------------------------------------------------------------------
>> Check out the vibrant tech community on one of the world's most
>> engaging tech sites, Slashdot.org!
https://urldefense.proofpoint.com/v2/url?u=http-3A__sdm.link_slashdot&d=DwIFaQ&c=RoP1YumCXCgaWHvlZYR8PQcxBKCX5YTpkKY057SbK10&r=Ua8D9LRbRt6_3fFyQghU61tppZpWmDKj9QGctDqRYQQ&m=cH3SiGpv68sRD_aUN_72YOkq1pGTpnhFAu7MR6YykPk&s=iauIH1J2V2la7WFYkj7B-Uy2a-VI8wEPtnMjxtg-hYI&e=
>> _______________________________________________
>> Opensaf-devel mailing list
>> Opensaf-devel@lists.sourceforge.net
>>
https://urldefense.proofpoint.com/v2/url?u=https-3A__lists.sourceforge.net_lists_listinfo_opensaf-2Ddevel&d=DwIFaQ&c=RoP1YumCXCgaWHvlZYR8PQcxBKCX5YTpkKY057SbK10&r=Ua8D9LRbRt6_3fFyQghU61tppZpWmDKj9QGctDqRYQQ&m=cH3SiGpv68sRD_aUN_72YOkq1pGTpnhFAu7MR6YykPk&s=DZ_Qgzmg4iAfECXSjqISBMDnnl4junMlXKEPECgKYM4&e=
>>
------------------------------------------------------------------------------
Check out the vibrant tech community on one of the world's most
engaging tech sites, Slashdot.org! http://sdm.link/slashdot
_______________________________________________
Opensaf-devel mailing list
Opensaf-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/opensaf-devel
