From: Chuck Lever <chuck.lever@oracle.com>
To: Trond Myklebust <trondmy@hammerspace.com>
Cc: Linux NFS Mailing List <linux-nfs@vger.kernel.org>
Subject: Re: [PATCH v3 26/44] SUNRPC: Improve latency for interactive tasks
Date: Wed, 2 Jan 2019 13:17:59 -0500 [thread overview]
Message-ID: <E9A190A9-4A55-4D02-9259-2F26BC41F16C@oracle.com> (raw)
In-Reply-To: <07dcc1731996d6e59d882c5e4e7e7765d013c337.camel@hammerspace.com>
> On Dec 31, 2018, at 2:21 PM, Trond Myklebust <trondmy@hammerspace.com> wrote:
>
> On Mon, 2018-12-31 at 19:18 +0000, Trond Myklebust wrote:
>> On Mon, 2018-12-31 at 14:09 -0500, Chuck Lever wrote:
>>>> On Dec 31, 2018, at 1:59 PM, Trond Myklebust <
>>>> trondmy@hammerspace.com> wrote:
>>>>
>>>> On Mon, 2018-12-31 at 13:44 -0500, Chuck Lever wrote:
>>>>>> On Dec 31, 2018, at 1:09 PM, Trond Myklebust <
>>>>>> trondmy@hammerspace.com> wrote:
>>>>>>
>>>>>> On Thu, 2018-12-27 at 17:34 -0500, Chuck Lever wrote:
>>>>>>>> On Dec 27, 2018, at 5:14 PM, Trond Myklebust <
>>>>>>>> trondmy@hammerspace.com> wrote:
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>> On Dec 27, 2018, at 20:21, Chuck Lever <
>>>>>>>>> chuck.lever@oracle.com>
>>>>>>>>> wrote:
>>>>>>>>>
>>>>>>>>> Hi Trond-
>>>>>>>>>
>>>>>>>>> I've chased down a couple of remaining regressions with
>>>>>>>>> the
>>>>>>>>> v4.20
>>>>>>>>> NFS client,
>>>>>>>>> and they seem to be rooted in this commit.
>>>>>>>>>
>>>>>>>>> When using sec=krb5, krb5i, or krb5p I found that
>>>>>>>>> multi-
>>>>>>>>> threaded
>>>>>>>>> workloads
>>>>>>>>> trigger a lot of server-side disconnects. This is with
>>>>>>>>> TCP
>>>>>>>>> and
>>>>>>>>> RDMA transports.
>>>>>>>>> An instrumented server shows that the client is under-
>>>>>>>>> running
>>>>>>>>> the
>>>>>>>>> GSS sequence
>>>>>>>>> number window. I monitored the order in which GSS
>>>>>>>>> sequence
>>>>>>>>> numbers appear on
>>>>>>>>> the wire, and after this commit, the sequence numbers
>>>>>>>>> are
>>>>>>>>> wildly
>>>>>>>>> misordered.
>>>>>>>>> If I revert the hunk in xprt_request_enqueue_transmit,
>>>>>>>>> the
>>>>>>>>> problem goes away.
>>>>>>>>>
>>>>>>>>> I also found that reverting that hunk results in a 3-4%
>>>>>>>>> improvement in fio
>>>>>>>>> IOPS rates, as well as improvement in average and
>>>>>>>>> maximum
>>>>>>>>> latency
>>>>>>>>> as reported
>>>>>>>>> by fio.
>>>>>>>>>
>>>>>>>>
>>>>>>>> Hmm… Provided the sequence numbers still lie within the
>>>>>>>> window,
>>>>>>>> then why would the order matter?
>>>>>>>
>>>>>>> The misordering is so bad that one request is delayed long
>>>>>>> enough
>>>>>>> to
>>>>>>> fall outside the window. The new “need re-encode” logic
>>>>>>> does
>>>>>>> not
>>>>>>> trigger.
>>>>>>>
>>>>>>
>>>>>> That's weird. I can't see anything wrong with need re-encode
>>>>>> at
>>>>>> this
>>>>>> point.
>>>>>
>>>>> I don't think there is anything wrong with it, it looks like
>>>>> it's
>>>>> not called in this case.
>>>>
>>>> So you are saying that the call to rpcauth_xmit_need_reencode()
>>>> is
>>>> triggering the EBADMSG, but that this fails to cause a re-encode
>>>> of
>>>> the
>>>> message?
>>>
>>> No, I think what's going on is that the need_reencode happens when
>>> the
>>> RPC is enqueued, and is successful.
>>>
>>> But xprt_request_enqueue_transmit places the RPC somewhere in the
>>> middle
>>> of xmit_queue. xmit_queue is long enough that more than 128
>>> requests
>>> are
>>> before the enqueued request.
>>
>> The test for rpcauth_xmit_need_reencode() happens when we call
>> xprt_request_transmit() to actually put the RPC call on the wire. The
>> enqueue order should not be able to defeat that test.
>>
>> Hmm... Is it perhaps the test for req->rq_bytes_sent that is failing
>> because this is a retransmission after a disconnect/reconnect that
>> didn't trigger a re-encode?
>
> Actually, it might be worth a try to move the test for
> rpcauth_xmit_need_reencode() outside the enclosing test for req-
>> rq_bytes_sent as that is just a minor optimisation.
Perhaps that's the case for TCP, but RPCs sent via xprtrdma never set
req->rq_bytes_sent to a non-zero value. The body of the "if" statement
is always executed for those RPCs.
--
Chuck Lever
next prev parent reply other threads:[~2019-01-02 18:18 UTC|newest]
Thread overview: 76+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-09-17 13:02 [PATCH v3 00/44] Convert RPC client transmission to a queued model Trond Myklebust
2018-09-17 13:02 ` [PATCH v3 01/44] SUNRPC: Clean up initialisation of the struct rpc_rqst Trond Myklebust
2018-09-17 13:02 ` [PATCH v3 02/44] SUNRPC: If there is no reply expected, bail early from call_decode Trond Myklebust
2018-09-17 13:02 ` [PATCH v3 03/44] SUNRPC: The transmitted message must lie in the RPCSEC window of validity Trond Myklebust
2018-09-17 13:02 ` [PATCH v3 04/44] SUNRPC: Simplify identification of when the message send/receive is complete Trond Myklebust
2018-09-17 13:02 ` [PATCH v3 05/44] SUNRPC: Avoid holding locks across the XDR encoding of the RPC message Trond Myklebust
2018-09-17 13:02 ` [PATCH v3 06/44] SUNRPC: Rename TCP receive-specific state variables Trond Myklebust
2018-09-17 13:02 ` [PATCH v3 07/44] SUNRPC: Move reset of TCP state variables into the reconnect code Trond Myklebust
2018-09-17 13:02 ` [PATCH v3 08/44] SUNRPC: Add socket transmit queue offset tracking Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 09/44] SUNRPC: Simplify dealing with aborted partially transmitted messages Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 10/44] SUNRPC: Refactor the transport request pinning Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 11/44] SUNRPC: Add a helper to wake up a sleeping rpc_task and set its status Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 12/44] SUNRPC: Test whether the task is queued before grabbing the queue spinlocks Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 13/44] SUNRPC: Don't wake queued RPC calls multiple times in xprt_transmit Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 14/44] SUNRPC: Rename xprt->recv_lock to xprt->queue_lock Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 15/44] SUNRPC: Refactor xprt_transmit() to remove the reply queue code Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 16/44] SUNRPC: Refactor xprt_transmit() to remove wait for reply code Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 17/44] SUNRPC: Minor cleanup for call_transmit() Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 18/44] SUNRPC: Distinguish between the slot allocation list and receive queue Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 19/44] SUNRPC: Add a transmission queue for RPC requests Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 20/44] SUNRPC: Refactor RPC call encoding Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 21/44] SUNRPC: Fix up the back channel transmit Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 22/44] SUNRPC: Treat the task and request as separate in the xprt_ops->send_request() Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 23/44] SUNRPC: Don't reset the request 'bytes_sent' counter when releasing XPRT_LOCK Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 24/44] SUNRPC: Simplify xprt_prepare_transmit() Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 25/44] SUNRPC: Move RPC retransmission stat counter to xprt_transmit() Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 26/44] SUNRPC: Improve latency for interactive tasks Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 27/44] SUNRPC: Support for congestion control when queuing is enabled Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 28/44] SUNRPC: Enqueue swapper tagged RPCs at the head of the transmit queue Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 29/44] SUNRPC: Allow calls to xprt_transmit() to drain the entire " Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 30/44] SUNRPC: Allow soft RPC calls to time out when waiting for the XPRT_LOCK Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 31/44] SUNRPC: Turn off throttling of RPC slots for TCP sockets Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 32/44] SUNRPC: Clean up transport write space handling Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 33/44] SUNRPC: Cleanup: remove the unused 'task' argument from the request_send() Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 34/44] SUNRPC: Don't take transport->lock unnecessarily when taking XPRT_LOCK Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 35/44] SUNRPC: Convert xprt receive queue to use an rbtree Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 36/44] SUNRPC: Fix priority queue fairness Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 37/44] SUNRPC: Convert the xprt->sending queue back to an ordinary wait queue Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 38/44] SUNRPC: Add a label for RPC calls that require allocation on receive Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 39/44] SUNRPC: Add a bvec array to struct xdr_buf for use with iovec_iter() Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 40/44] SUNRPC: Simplify TCP receive code by switching to using iterators Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 41/44] SUNRPC: Clean up - rename xs_tcp_data_receive() to xs_stream_data_receive() Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 42/44] SUNRPC: Allow AF_LOCAL sockets to use the generic stream receive Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 43/44] SUNRPC: Clean up xs_udp_data_receive() Trond Myklebust
2018-09-17 13:03 ` [PATCH v3 44/44] SUNRPC: Unexport xdr_partial_copy_from_skb() Trond Myklebust
2018-09-17 20:44 ` [PATCH v3 40/44] SUNRPC: Simplify TCP receive code by switching to using iterators Trond Myklebust
2018-11-09 11:19 ` Catalin Marinas
2018-11-29 19:28 ` Cristian Marussi
2018-11-29 19:56 ` Trond Myklebust
2018-11-30 16:19 ` Cristian Marussi
2018-11-30 19:31 ` Trond Myklebust
2018-12-02 16:44 ` Trond Myklebust
2018-12-03 11:45 ` Catalin Marinas
2018-12-03 11:53 ` Cristian Marussi
2018-12-03 18:54 ` Cristian Marussi
2018-12-27 19:21 ` [PATCH v3 26/44] SUNRPC: Improve latency for interactive tasks Chuck Lever
2018-12-27 22:14 ` Trond Myklebust
2018-12-27 22:34 ` Chuck Lever
2018-12-31 18:09 ` Trond Myklebust
2018-12-31 18:44 ` Chuck Lever
2018-12-31 18:59 ` Trond Myklebust
2018-12-31 19:09 ` Chuck Lever
2018-12-31 19:18 ` Trond Myklebust
2018-12-31 19:21 ` Trond Myklebust
2019-01-02 18:17 ` Chuck Lever [this message]
2019-01-02 18:45 ` Trond Myklebust
2019-01-02 18:51 ` Chuck Lever
2019-01-02 18:57 ` Trond Myklebust
2019-01-02 19:06 ` Trond Myklebust
2019-01-02 19:24 ` Trond Myklebust
2019-01-02 19:33 ` Chuck Lever
2019-01-02 19:08 ` Chuck Lever
2019-01-02 19:11 ` Trond Myklebust
2018-09-18 21:01 ` [PATCH v3 15/44] SUNRPC: Refactor xprt_transmit() to remove the reply queue code Anna Schumaker
2018-09-19 15:48 ` Trond Myklebust
2018-09-19 17:30 ` Anna Schumaker
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=E9A190A9-4A55-4D02-9259-2F26BC41F16C@oracle.com \
--to=chuck.lever@oracle.com \
--cc=linux-nfs@vger.kernel.org \
--cc=trondmy@hammerspace.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).