qemu-block
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH 1/1] nbd: trace long NBD operations


From: Vladimir Sementsov-Ogievskiy
Subject: Re: [PATCH 1/1] nbd: trace long NBD operations
Date: Fri, 27 May 2022 12:33:36 +0300
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101 Thunderbird/91.8.1

On 5/27/22 11:43, Denis V. Lunev wrote:
At the moment there are 2 sources of lengthy operations if configured:
* open connection, which could retry inside and
* reconnect of already opened connection
These operations could be quite lengthy and cumbersome to catch thus
it would be quite natural to add trace points for them.

This patch is based on the original downstream work made by Vladimir.

Signed-off-by: Denis V. Lunev <den@openvz.org>
CC: Eric Blake <eblake@redhat.com>
CC: Vladimir Sementsov-Ogievskiy <vsementsov@yandex-team.ru>
CC: Kevin Wolf <kwolf@redhat.com>
CC: Hanna Reitz <hreitz@redhat.com>
CC: Paolo Bonzini <pbonzini@redhat.com>
---
  block/nbd.c             | 11 ++++++++---
  block/trace-events      |  2 ++
  nbd/client-connection.c |  2 ++
  nbd/trace-events        |  3 +++
  4 files changed, 15 insertions(+), 3 deletions(-)

diff --git a/block/nbd.c b/block/nbd.c
index 6085ab1d2c..f1a473d36b 100644
--- a/block/nbd.c
+++ b/block/nbd.c
@@ -371,6 +371,7 @@ static bool nbd_client_connecting(BDRVNBDState *s)
  /* Called with s->requests_lock taken.  */
  static coroutine_fn void nbd_reconnect_attempt(BDRVNBDState *s)
  {
+    int ret;
      bool blocking = s->state == NBD_CLIENT_CONNECTING_WAIT;
/*
@@ -380,6 +381,8 @@ static coroutine_fn void nbd_reconnect_attempt(BDRVNBDState 
*s)
      assert(nbd_client_connecting(s));
      assert(s->in_flight == 1);
+ trace_nbd_reconnect_attempt(s->bs->in_flight);
+
      if (blocking && !s->reconnect_delay_timer) {
          /*
           * It's the first reconnect attempt after switching to
@@ -401,7 +404,7 @@ static coroutine_fn void nbd_reconnect_attempt(BDRVNBDState 
*s)
      }
qemu_mutex_unlock(&s->requests_lock);
-    nbd_co_do_establish_connection(s->bs, blocking, NULL);
+    ret = nbd_co_do_establish_connection(s->bs, blocking, NULL);
      qemu_mutex_lock(&s->requests_lock);
/*
@@ -410,6 +413,8 @@ static coroutine_fn void nbd_reconnect_attempt(BDRVNBDState 
*s)
       * this I/O request (so draining removes all timers).
       */
      reconnect_delay_timer_del(s);
+
+    trace_nbd_reconnect_attempt_result(ret, s->bs->in_flight);

May be better trace exactly after nbd_co_do_establish_connection(). Doesn't 
really matter, just simpler code.

  }
static coroutine_fn int nbd_receive_replies(BDRVNBDState *s, uint64_t handle)
@@ -1856,8 +1861,8 @@ static int nbd_process_options(BlockDriverState *bs, 
QDict *options,
          goto error;
      }
- s->reconnect_delay = qemu_opt_get_number(opts, "reconnect-delay", 0);
-    s->open_timeout = qemu_opt_get_number(opts, "open-timeout", 0);
+    s->reconnect_delay = qemu_opt_get_number(opts, "reconnect-delay", 300);
+    s->open_timeout = qemu_opt_get_number(opts, "open-timeout", 300);

That's changing defaults. Should not be in this patch. And I don't think we can 
simply change upstream default of open-timeout, as it breaks habitual behavior.

ret = 0; diff --git a/block/trace-events b/block/trace-events
index 549090d453..caab699c22 100644
--- a/block/trace-events
+++ b/block/trace-events
@@ -172,6 +172,8 @@ nbd_read_reply_entry_fail(int ret, const char *err) "ret = %d, 
err: %s"
  nbd_co_request_fail(uint64_t from, uint32_t len, uint64_t handle, uint16_t flags, uint16_t type, const char *name, int ret, const char 
*err) "Request failed { .from = %" PRIu64", .len = %" PRIu32 ", .handle = %" PRIu64 ", .flags = 
0x%" PRIx16 ", .type = %" PRIu16 " (%s) } ret = %d, err: %s"
  nbd_client_handshake(const char *export_name) "export '%s'"
  nbd_client_handshake_success(const char *export_name) "export '%s'"
+nbd_reconnect_attempt(int in_flight) "in_flight %d"
+nbd_reconnect_attempt_result(int ret, int in_flight) "ret %d in_flight %d"

bs->in_flight is "unsigned int", so a bit better would be use "unsigned int" and 
"%u" here

# ssh.c
  ssh_restart_coroutine(void *co) "co=%p"
diff --git a/nbd/client-connection.c b/nbd/client-connection.c
index 2a632931c3..a5ee82e804 100644
--- a/nbd/client-connection.c
+++ b/nbd/client-connection.c
@@ -23,6 +23,7 @@
   */
#include "qemu/osdep.h"
+#include "trace.h"
#include "block/nbd.h" @@ -210,6 +211,7 @@ static void *connect_thread_func(void *opaque)
              object_unref(OBJECT(conn->sioc));
              conn->sioc = NULL;
              if (conn->do_retry && !conn->detached) {
+                trace_nbd_connect_iteration(timeout);

Here we are going to sleep a bit, before next reconnect attempt. I'd call the trace-point 
"trace_nbd_connect_thread_sleep" or something like this to be more intuitive.

                  qemu_mutex_unlock(&conn->mutex);
sleep(timeout);
diff --git a/nbd/trace-events b/nbd/trace-events
index c4919a2dd5..bdadfdc82d 100644
--- a/nbd/trace-events
+++ b/nbd/trace-events
@@ -73,3 +73,6 @@ nbd_co_receive_request_decode_type(uint64_t handle, uint16_t 
type, const char *n
  nbd_co_receive_request_payload_received(uint64_t handle, uint32_t len) "Payload received: 
handle = %" PRIu64 ", len = %" PRIu32
  nbd_co_receive_align_compliance(const char *op, uint64_t from, uint32_t len, uint32_t align) "client 
sent non-compliant unaligned %s request: from=0x%" PRIx64 ", len=0x%" PRIx32 ", 
align=0x%" PRIx32
  nbd_trip(void) "Reading request"
+
+# client-connection.c
+nbd_connect_iteration(int in_flight) "timeout %d"


--
Best regards,
Vladimir



reply via email to

[Prev in Thread] Current Thread [Next in Thread]