xref: /linux/Documentation/networking/ipvs-sysctl.rst (revision 05e352444b2430de4b183b4a988085381e5fd6ad)
1.. SPDX-License-Identifier: GPL-2.0
2
3===========
4IPvs-sysctl
5===========
6
7/proc/sys/net/ipv4/vs/* Variables:
8==================================
9
10am_droprate - INTEGER
11	default 10
12
13	It sets the always mode drop rate, which is used in the mode 3
14	of the drop_rate defense.
15
16amemthresh - INTEGER
17	default 1024
18
19	It sets the available memory threshold (in pages), which is
20	used in the automatic modes of defense. When there is no
21	enough available memory, the respective strategy will be
22	enabled and the variable is automatically set to 2, otherwise
23	the strategy is disabled and the variable is  set  to 1.
24
25backup_only - BOOLEAN
26	- 0 - disabled (default)
27	- not 0 - enabled
28
29	If set, disable the director function while the server is
30	in backup mode to avoid packet loops for DR/TUN methods.
31
32conn_lfactor - INTEGER
33	Possible values: -8 (larger table) .. 8 (smaller table)
34
35	Default: -4
36
37	Controls the sizing of the connection hash table based on the
38	load factor (number of connections per table buckets):
39
40		2^conn_lfactor = nodes / buckets
41
42	As result, the table grows if load increases and shrinks when
43	load decreases in the range of 2^8 - 2^conn_tab_bits (module
44	parameter).
45	The value is a shift count where negative values select
46	buckets = (connection hash nodes << -value) while positive
47	values select buckets = (connection hash nodes >> value). The
48	negative values reduce the collisions and reduce the time for
49	lookups but increase the table size. Positive values will
50	tolerate load above 100% when using smaller table is
51	preferred with the cost of more collisions. If using NAT
52	connections consider decreasing the value with one because
53	they add two nodes in the hash table.
54
55	Example:
56	-4: grow if load goes above 6% (buckets = nodes * 16)
57	2: grow if load goes above 400% (buckets = nodes / 4)
58
59conn_reuse_mode - INTEGER
60	1 - default
61
62	Controls how ipvs will deal with connections that are detected
63	port reuse. It is a bitmap, with the values being:
64
65	0: disable any special handling on port reuse. The new
66	connection will be delivered to the same real server that was
67	servicing the previous connection.
68
69	bit 1: enable rescheduling of new connections when it is safe.
70	That is, whenever expire_nodest_conn and for TCP sockets, when
71	the connection is in TIME_WAIT state (which is only possible if
72	you use NAT mode).
73
74	bit 2: it is bit 1 plus, for TCP connections, when connections
75	are in FIN_WAIT state, as this is the last state seen by load
76	balancer in Direct Routing mode. This bit helps on adding new
77	real servers to a very busy cluster.
78
79conntrack - BOOLEAN
80	- 0 - disabled (default)
81	- not 0 - enabled
82
83	If set, maintain connection tracking entries for
84	connections handled by IPVS.
85
86	This should be enabled if connections handled by IPVS are to be
87	also handled by stateful firewall rules. That is, iptables rules
88	that make use of connection tracking.  It is a performance
89	optimisation to disable this setting otherwise.
90
91	Connections handled by the IPVS FTP application module
92	will have connection tracking entries regardless of this setting.
93
94	Only available when IPVS is compiled with CONFIG_IP_VS_NFCT enabled.
95
96cache_bypass - BOOLEAN
97	- 0 - disabled (default)
98	- not 0 - enabled
99
100	If it is enabled, forward packets to the original destination
101	directly when no cache server is available and destination
102	address is not local (iph->daddr is RTN_UNICAST). It is mostly
103	used in transparent web cache cluster.
104
105debug_level - INTEGER
106	- 0          - transmission error messages (default)
107	- 1          - non-fatal error messages
108	- 2          - configuration
109	- 3          - destination trash
110	- 4          - drop entry
111	- 5          - service lookup
112	- 6          - scheduling
113	- 7          - connection new/expire, lookup and synchronization
114	- 8          - state transition
115	- 9          - binding destination, template checks and applications
116	- 10         - IPVS packet transmission
117	- 11         - IPVS packet handling (ip_vs_in/ip_vs_out)
118	- 12 or more - packet traversal
119
120	Only available when IPVS is compiled with CONFIG_IP_VS_DEBUG enabled.
121
122	Higher debugging levels include the messages for lower debugging
123	levels, so setting debug level 2, includes level 0, 1 and 2
124	messages. Thus, logging becomes more and more verbose the higher
125	the level.
126
127drop_entry - INTEGER
128	- 0  - disabled (default)
129
130	The drop_entry defense is to randomly drop entries in the
131	connection hash table, just in order to collect back some
132	memory for new connections. In the current code, the
133	drop_entry procedure can be activated every second, then it
134	randomly scans 1/32 of the whole and drops entries that are in
135	the SYN-RECV/SYNACK state, which should be effective against
136	syn-flooding attack.
137
138	The valid values of drop_entry are from 0 to 3, where 0 means
139	that this strategy is always disabled, 1 and 2 mean automatic
140	modes (when there is no enough available memory, the strategy
141	is enabled and the variable is automatically set to 2,
142	otherwise the strategy is disabled and the variable is set to
143	1), and 3 means that the strategy is always enabled.
144
145drop_packet - INTEGER
146	- 0  - disabled (default)
147
148	The drop_packet defense is designed to drop 1/rate packets
149	before forwarding them to real servers. If the rate is 1, then
150	drop all the incoming packets.
151
152	The value definition is the same as that of the drop_entry. In
153	the automatic mode, the rate is determined by the follow
154	formula: rate = amemthresh / (amemthresh - available_memory)
155	when available memory is less than the available memory
156	threshold. When the mode 3 is set, the always mode drop rate
157	is controlled by the /proc/sys/net/ipv4/vs/am_droprate.
158
159est_cpulist - CPULIST
160	Allowed	CPUs for estimation kthreads
161
162	Syntax: standard cpulist format
163	empty list - stop kthread tasks and estimation
164	default - the system's housekeeping CPUs for kthreads
165
166	Example:
167	"all": all possible CPUs
168	"0-N": all possible CPUs, N denotes last CPU number
169	"0,1-N:1/2": first and all CPUs with odd number
170	"": empty list
171
172est_nice - INTEGER
173	default 0
174	Valid range: -20 (more favorable) .. 19 (less favorable)
175
176	Niceness value to use for the estimation kthreads (scheduling
177	priority)
178
179expire_nodest_conn - BOOLEAN
180	- 0 - disabled (default)
181	- not 0 - enabled
182
183	The default value is 0, the load balancer will silently drop
184	packets when its destination server is not available. It may
185	be useful, when user-space monitoring program deletes the
186	destination server (because of server overload or wrong
187	detection) and add back the server later, and the connections
188	to the server can continue.
189
190	If this feature is enabled, the load balancer will expire the
191	connection immediately when a packet arrives and its
192	destination server is not available, then the client program
193	will be notified that the connection is closed. This is
194	equivalent to the feature some people requires to flush
195	connections when its destination is not available.
196
197expire_quiescent_template - BOOLEAN
198	- 0 - disabled (default)
199	- not 0 - enabled
200
201	When set to a non-zero value, the load balancer will expire
202	persistent templates when the destination server is quiescent.
203	This may be useful, when a user makes a destination server
204	quiescent by setting its weight to 0 and it is desired that
205	subsequent otherwise persistent connections are sent to a
206	different destination server.  By default new persistent
207	connections are allowed to quiescent destination servers.
208
209	If this feature is enabled, the load balancer will expire the
210	persistence template if it is to be used to schedule a new
211	connection and the destination server is quiescent.
212
213ignore_tunneled - BOOLEAN
214	- 0 - disabled (default)
215	- not 0 - enabled
216
217	If set, ipvs will set the ipvs_property on all packets which are of
218	unrecognized protocols.  This prevents us from routing tunneled
219	protocols like ipip, which is useful to prevent rescheduling
220	packets that have been tunneled to the ipvs host (i.e. to prevent
221	ipvs routing loops when ipvs is also acting as a real server).
222
223nat_icmp_send - BOOLEAN
224	- 0 - disabled (default)
225	- not 0 - enabled
226
227	It controls sending icmp error messages (ICMP_DEST_UNREACH)
228	for VS/NAT when the load balancer receives packets from real
229	servers but the connection entries don't exist.
230
231pmtu_disc - BOOLEAN
232	- 0 - disabled
233	- not 0 - enabled (default)
234
235	By default, reject with FRAG_NEEDED all DF packets that exceed
236	the PMTU, irrespective of the forwarding method. For TUN method
237	the flag can be disabled to fragment such packets.
238
239secure_tcp - INTEGER
240	- 0  - disabled (default)
241
242	The secure_tcp defense is to use a more complicated TCP state
243	transition table. For VS/NAT, it also delays entering the
244	TCP ESTABLISHED state until the three way handshake is completed.
245
246	The value definition is the same as that of drop_entry and
247	drop_packet.
248
249svc_lfactor - INTEGER
250	Possible values: -8 (larger table) .. 8 (smaller table)
251
252	Default: -3
253
254	Controls the sizing of the service hash table based on the
255	load factor (number of services per table buckets). The table
256	will grow and shrink in the range of 2^4 - 2^20.
257	See conn_lfactor for explanation.
258
259sync_threshold - vector of 2 INTEGERs: sync_threshold, sync_period
260	default 3 50
261
262	It sets synchronization threshold, which is the minimum number
263	of incoming packets that a connection needs to receive before
264	the connection will be synchronized. A connection will be
265	synchronized, every time the number of its incoming packets
266	modulus sync_period equals the threshold. The range of the
267	threshold is from 0 to sync_period.
268
269	When sync_period and sync_refresh_period are 0, send sync only
270	for state changes or only once when pkts matches sync_threshold
271
272sync_refresh_period - UNSIGNED INTEGER
273	default 0
274
275	In seconds, difference in reported connection timer that triggers
276	new sync message. It can be used to avoid sync messages for the
277	specified period (or half of the connection timeout if it is lower)
278	if connection state is not changed since last sync.
279
280	This is useful for normal connections with high traffic to reduce
281	sync rate. Additionally, retry sync_retries times with period of
282	sync_refresh_period/8.
283
284sync_retries - INTEGER
285	default 0
286
287	Defines sync retries with period of sync_refresh_period/8. Useful
288	to protect against loss of sync messages. The range of the
289	sync_retries is from 0 to 3.
290
291sync_qlen_max - UNSIGNED LONG
292
293	Hard limit for queued sync messages that are not sent yet. It
294	defaults to 1/32 of the memory pages but actually represents
295	number of messages. It will protect us from allocating large
296	parts of memory when the sending rate is lower than the queuing
297	rate.
298
299sync_sock_size - INTEGER
300	default 0
301
302	Configuration of SNDBUF (master) or RCVBUF (slave) socket limit.
303	Default value is 0 (preserve system defaults).
304
305sync_ports - INTEGER
306	default 1
307
308	The number of threads that master and backup servers can use for
309	sync traffic. Every thread will use single UDP port, thread 0 will
310	use the default port 8848 while last thread will use port
311	8848+sync_ports-1.
312
313snat_reroute - BOOLEAN
314	- 0 - disabled
315	- not 0 - enabled (default)
316
317	If enabled, recalculate the route of SNATed packets from
318	realservers so that they are routed as if they originate from the
319	director. Otherwise they are routed as if they are forwarded by the
320	director.
321
322	If policy routing is in effect then it is possible that the route
323	of a packet originating from a director is routed differently to a
324	packet being forwarded by the director.
325
326	If policy routing is not in effect then the recalculated route will
327	always be the same as the original route so it is an optimisation
328	to disable snat_reroute and avoid the recalculation.
329
330sync_persist_mode - INTEGER
331	default 0
332
333	Controls the synchronisation of connections when using persistence
334
335	0: All types of connections are synchronised
336
337	1: Attempt to reduce the synchronisation traffic depending on
338	the connection type. For persistent services avoid synchronisation
339	for normal connections, do it only for persistence templates.
340	In such case, for TCP and SCTP it may need enabling sloppy_tcp and
341	sloppy_sctp flags on backup servers. For non-persistent services
342	such optimization is not applied, mode 0 is assumed.
343
344sync_version - INTEGER
345	default 1
346
347	The version of the synchronisation protocol used when sending
348	synchronisation messages.
349
350	0 selects the original synchronisation protocol (version 0). This
351	should be used when sending synchronisation messages to a legacy
352	system that only understands the original synchronisation protocol.
353
354	1 selects the current synchronisation protocol (version 1). This
355	should be used where possible.
356
357	Kernels with this sync_version entry are able to receive messages
358	of both version 1 and version 2 of the synchronisation protocol.
359
360run_estimation - BOOLEAN
361	0 - disabled
362	not 0 - enabled (default)
363
364	If disabled, the estimation will be suspended and kthread tasks
365	stopped.
366
367	You can always re-enable estimation by setting this value to 1.
368	But be careful, the first estimation after re-enable is not
369	accurate.
370