| Dead Connection Detection (DCD) Explained [ID 151972.1]
|
|
| |
修改時間
17-AUG-2009 類型
BULLETIN 狀態
PUBLISHED |
|
Checked for relevance on 5-FEB-2009.
DEAD CONNECTION DETECTION
=========================
OVERVIEW
--------
Dead Connection Detection (DCD) is a feature of SQL*Net 2.1 and later, including
Oracle Net8 and Oracle NET. DCD detects when a partner in a SQL*Net V2 client/server
or server/server connection has terminated unexpectedly, and flags the dead session
so PMON can release the resources associated with it.
DCD is intended primarily for environments in which clients power down their
systems without disconnecting from their Oracle sessions, a problem
characteristic of networks with PC clients.
DCD is initiated on the server when a connection is established. At this
time SQL*Net reads the SQL*Net parameter files and sets a timer to generate an
alarm. The timer interval is set by providing a non-zero value in minutes for
the SQLNET.EXPIRE_TIME parameter in the sqlnet.ora file. The Database and Listener
need to be restarted after any DCD changes.
When the timer expires, SQL*Net on the server sends a "probe" packet to the
client. (In the case of a database link, the destination of the link
constitutes the server side of the connection.) The probe is essentially an
empty SQL*Net packet and does not represent any form of SQL*Net level data,
but it creates data traffic on the underlying protocol.
If the client end of the connection is still active, the probe is discarded,
and the timer mechanism is reset. If the client has terminated abnormally,
the server will receive an error from the send call issued for the probe, and
SQL*Net on the server will signal the operating system to release the
connection's resources.
On Unix servers, the sqlnet.ora file must be in either $TNS_ADMIN or
$ORACLE_HOME/network/admin. Neither /etc nor /var/opt/oracle alone is valid.
It should be also be noted that in SQL*Net 2.1.x, an active orphan process
(one processing a query, for example) will not be killed until the query
completes. In SQL*Net 2.2, orphaned resources will be released regardless of
activity.
This is a server feature only. The client may be running any supported
SQL*Net V2 release.
THE FUNCTION OF THE PROTOCOL STACK
----------------------------------
While Dead Connection Detection is set at the SQL*Net level, it relies heavily
on the underlying protocol stack for it's successful execution. For example,
you might set SQLNET.EXPIRE_TIME=1 in the sqlnet.ora file, but it is unlikely
that an orphaned server process will be cleaned up immediately upon expiration
of that interval.
TCP/IP, for example, is a connection-oriented protocol, and as such, the
protocol will implement some level of packet timeout and retransmission in an
effort to guarantee the safe and sequenced order of data packets. If a timely
acknowledgement is not received in response to the probe packet, the TCP/IP
stack will retransmit the packet some number of times before timing out. After
TCP/IP gives up, then SQL*Net receives notification that the probe failed.
The time that it takes TCP/IP to timeout is dependent on the TCP/IP stack, and
timeouts of many minutes are entirely common. This has been an area of concern
for many customers, as many retransmissions at the protocol layer causes what
could be a significant lag between the expiration of the DCD interval and the
time when the orphaned process is actually killed.
The easiest way to determine if the protocol stack is causing such a delay
involves testing different DCD intervals.
TESTING THE PROTOCOL STACK
--------------------------
Set the SQLNET.EXPIRE_TIME parameter to 1 minute and note the time required to
clean up an orphaned server process. Then set SQLNET.EXPIRE_TIME to 5 minutes
and again observe the time required to clean up the shadow. If the TCP/IP
timeout is the reason the server resources do not get released, the time to
clean up the shadow should increase by about 4 minutes.
If the TCP/IP retransmission timeout is indeed the problem, the Operating
System kernel can be tuned to reduce the interval for and number of packet
retransmissions (on many Unix platforms, the file
/usr/include/netinet/tcp_timer.h contains the configuration parameters).
Reducing the interval and number of retransmissions may impact other system
components, since in effect you are shrinking the window allowed for
connections to process data, possibly resulting in inadvertent loss of
connections during periods of heavy system load. Slower connections from
remote sites may be impacted by this change.
Kernel parameters that may affect retransmission include but are not limited
to TCP_TTL, TCPTV_PERSMIN, TCPTV_MAX, and TCP_LINGERTIME.
*** To avoid disrupting other system processes, it is important to contact the
appropriate vendor for assistance in tuning the operating system kernel or
protocol stack. ***
MONITORING DEAD CONNECTION DETECTION
------------------------------------
The best way to determine if DCD is enabled and functioning properly is to
generate a server trace and search the file for the DCD probe packet. To
generate a server trace, set TRACE_LEVEL_SERVER=16 and
TRACE_DIRECTORY_SERVER=<path> in sqlnet.ora on the server (note the location
of the sqlnet.ora file). The resulting trace file will have a filename of
svr_<PID>.trc and will be located in the specified directory.
Is DCD Enabled?
---------------
For pre-Oracle8i versions, enable level 16 SQL*Net server tracing and search
the resultant server trace file for an entry like the following:
osntns: Enabling dead connection detection (1 min)
The timer interval listed should match the value of SQLNET.EXPIRE_TIME.
For Oracle8i onwards, you should see the following:
nstimini: entry
nstimig: entry
nstimig: normal exit
nstimini: initializing NLTM in asynchronous mode
nstimini: normal exit
nstimstart: entry
Is DCD Working?
---------------
Search the server trace file for DCD probe packets. They will appear in the
form of empty data packets, as follows:
nstimexp: entry
nstimexp: timer expired at 05-OCT-95 12:15:05
nsdo: entry
nsdo: cid=0, opcode=67, *bl=0, *what=1, uflgs=0x2, cflgs=0x3
nsdo: nsctx: state=8, flg=0x621c, mvd=0
nsdo: gtn=93, gtc=93, ptn=10, ptc=2048
nsdoacts: entry
nsdofls: entry
nsdofls: DATA flags: 0x0
nsdofls: sending NSPTDA packet
nspsend: entry
nspsend: plen=10, type=6
nttwr: entry
nttwr: socket 4 had bytes written=10
nttwr: exit
nspsend: 10 bytes to transport
nspsend:packet dump
nspsend:00 0A 00 00 06 00 00 00 |........|
nspsend:00 00 00 00 00 00 00 00 |........|
nspsend: normal exit
nsdofls: exit (0)
nsdoacts: flushing transport
nttctl: entry
nsdoacts: normal exit
nsdo: normal exit
nstimexp: normal exit
The entry:
nspsend:00 0A 00 00 06 00 00 00 |........|
nspsend:00 00 00 00 00 00 00 00 |........|
represents the probe packet. Note that DCD packets are 10 bytes long when they
are issued to the protocol stack. Once the protocol header and trailer bytes
for the underlying protocols have been added, the packet could be approximately
70 bytes long.
If DCD is enabled, you will see these probe packets written to the trace file
when the timer expires. If the server is a UNIX system, it might be useful to
establish a connection and tail the trace file:
tail -f svr_<PID>.trc
The time elapsed after each probe packet is written to the server trace should
match the SQLNET.EXPIRE_TIME value.
Note: from version 9.2.0.4.0 onwards, DCD probe packets are no longer traced in
SQL*Net trace files, however DCD packets can be observed using other forms of
tracing, such as network sniffer tracing.
KNOWN PROBLEMS OR LIMITATIONS
-----------------------------
- Of the few reported problems, perhaps the most significant is DCD's poor
performance on Windows NT. Dead connections are cleaned up only when the
server is rebooted and the database is restarted. Exactly how well DCD works
on NT depends on the client's protocol implementation. SQL*Net v2.3 has
improved the performance over earlier releases.
This has been logged as port-specific Bug#303578.
- On SCO Unix, a problem was reported in which server processes spin, consuming
large amounts of CPU, once the DCD timer expires. The problem is due to improper
signal handling and can be eliminated by disabling DCD.
This is port-specific Bug#293264
- Orphaned resources are not released if only the client application is
terminated. Only after the client PC has been rebooted does DCD release these
resources. For example, if a Windows application is killed yet Windows remains
running, the probe packet may be received and discarded as if the connection is
still active. As it currently stands, it appears that DCD detects dead client
machines, but not dead client processes.
This is logged as generic Bug#280848.
- The SQL*Net V2 implementation on MVS does not use the generic DCD mechanism,
and therefore the SQLNET.EXPIRE_TIME parameter does not apply. The KEEPALIVE
function of IBM's TCP/IP is used instead. This was implemented prior to
development of DCD.
This is documented in port-specific Bug#301318.
- DCD relies heavily on issuing probe packets during any phase of the connection.
This is not be possible with some protocols which run half-duplex. Hence, DCD is
not enabled on protocols like APPC/LU6.2.
This is not a bug, but is rather the intended design.
- Local connections using BEQ protocol adapters are not supported with DCD.
Local connections using the IPC protocol adapters are supported with DCD.
-BUG#1388806 : On Windows NT, DCD FAILS AFTER 16 CONNECTIONS
A FINAL NOTE...
--------------
On most OS'es (even more recent versions of Windows) if a process exits
abnormally or is killed by an administrator, the OS will still gracefully
clean up resources associated with that process including the network
connection(s). It will tell the server on the other end that it is closing
the network connection. DCD is still useful for times when there are problems
with the physical network (e.g. ethernet cable falls off the machine) or if
the OS kernel panics and crashes (e.g. blue screen of death) before it can
close the network connections. It may have another side benefit with certain
load balancing hardware, that may prematurely abort connections it thinks have
been idle too long, by sending a dummy packet to the client periodically.
Under no circumstances should you rely 100% on Dead Connection Detection.
It was developed to handle clients that have abnormally exited. Clients should
always exit their applications gracefully. It is the responsibility of the
application developer to make this possible. DCD is intended only to clean up
after abnormal events.
DCD is much more resource-intensive than similar mechanisms at the protocol
level, so if you depend on DCD to clean up all dead processes, that will put
an undue load on the server.
Clearly it is advantageous to exit applications cleanly in the first place.
