1 /proc/sys/net/ipv4/vs/* Variables:
6 It sets the always mode drop rate, which is used in the mode 3
7 of the drop_rate defense.
12 It sets the available memory threshold (in pages), which is
13 used in the automatic modes of defense. When there is no
14 enough available memory, the respective strategy will be
15 enabled and the variable is automatically set to 2, otherwise
16 the strategy is disabled and the variable is set to 1.
19 0 - disabled (default)
22 If set, disable the director function while the server is
23 in backup mode to avoid packet loops for DR/TUN methods.
25 conn_reuse_mode - INTEGER
28 Controls how ipvs will deal with connections that are detected
29 port reuse. It is a bitmap, with the values being:
31 0: disable any special handling on port reuse. The new
32 connection will be delivered to the same real server that was
33 servicing the previous connection.
35 bit 1: enable rescheduling of new connections when it is safe.
36 That is, whenever expire_nodest_conn and for TCP sockets, when
37 the connection is in TIME_WAIT state (which is only possible if
40 bit 2: it is bit 1 plus, for TCP connections, when connections
41 are in FIN_WAIT state, as this is the last state seen by load
42 balancer in Direct Routing mode. This bit helps on adding new
43 real servers to a very busy cluster.
46 0 - disabled (default)
49 If set, maintain connection tracking entries for
50 connections handled by IPVS.
52 This should be enabled if connections handled by IPVS are to be
53 also handled by stateful firewall rules. That is, iptables rules
54 that make use of connection tracking. It is a performance
55 optimisation to disable this setting otherwise.
57 Connections handled by the IPVS FTP application module
58 will have connection tracking entries regardless of this setting.
60 Only available when IPVS is compiled with CONFIG_IP_VS_NFCT enabled.
62 cache_bypass - BOOLEAN
63 0 - disabled (default)
66 If it is enabled, forward packets to the original destination
67 directly when no cache server is available and destination
68 address is not local (iph->daddr is RTN_UNICAST). It is mostly
69 used in transparent web cache cluster.
72 0 - transmission error messages (default)
73 1 - non-fatal error messages
79 7 - connection new/expire, lookup and synchronization
81 9 - binding destination, template checks and applications
82 10 - IPVS packet transmission
83 11 - IPVS packet handling (ip_vs_in/ip_vs_out)
84 12 or more - packet traversal
86 Only available when IPVS is compiled with CONFIG_IP_VS_DEBUG enabled.
88 Higher debugging levels include the messages for lower debugging
89 levels, so setting debug level 2, includes level 0, 1 and 2
90 messages. Thus, logging becomes more and more verbose the higher
94 0 - disabled (default)
96 The drop_entry defense is to randomly drop entries in the
97 connection hash table, just in order to collect back some
98 memory for new connections. In the current code, the
99 drop_entry procedure can be activated every second, then it
100 randomly scans 1/32 of the whole and drops entries that are in
101 the SYN-RECV/SYNACK state, which should be effective against
104 The valid values of drop_entry are from 0 to 3, where 0 means
105 that this strategy is always disabled, 1 and 2 mean automatic
106 modes (when there is no enough available memory, the strategy
107 is enabled and the variable is automatically set to 2,
108 otherwise the strategy is disabled and the variable is set to
109 1), and 3 means that that the strategy is always enabled.
111 drop_packet - INTEGER
112 0 - disabled (default)
114 The drop_packet defense is designed to drop 1/rate packets
115 before forwarding them to real servers. If the rate is 1, then
116 drop all the incoming packets.
118 The value definition is the same as that of the drop_entry. In
119 the automatic mode, the rate is determined by the follow
120 formula: rate = amemthresh / (amemthresh - available_memory)
121 when available memory is less than the available memory
122 threshold. When the mode 3 is set, the always mode drop rate
123 is controlled by the /proc/sys/net/ipv4/vs/am_droprate.
125 expire_nodest_conn - BOOLEAN
126 0 - disabled (default)
129 The default value is 0, the load balancer will silently drop
130 packets when its destination server is not available. It may
131 be useful, when user-space monitoring program deletes the
132 destination server (because of server overload or wrong
133 detection) and add back the server later, and the connections
134 to the server can continue.
136 If this feature is enabled, the load balancer will expire the
137 connection immediately when a packet arrives and its
138 destination server is not available, then the client program
139 will be notified that the connection is closed. This is
140 equivalent to the feature some people requires to flush
141 connections when its destination is not available.
143 expire_quiescent_template - BOOLEAN
144 0 - disabled (default)
147 When set to a non-zero value, the load balancer will expire
148 persistent templates when the destination server is quiescent.
149 This may be useful, when a user makes a destination server
150 quiescent by setting its weight to 0 and it is desired that
151 subsequent otherwise persistent connections are sent to a
152 different destination server. By default new persistent
153 connections are allowed to quiescent destination servers.
155 If this feature is enabled, the load balancer will expire the
156 persistence template if it is to be used to schedule a new
157 connection and the destination server is quiescent.
159 ignore_tunneled - BOOLEAN
160 0 - disabled (default)
163 If set, ipvs will set the ipvs_property on all packets which are of
164 unrecognized protocols. This prevents us from routing tunneled
165 protocols like ipip, which is useful to prevent rescheduling
166 packets that have been tunneled to the ipvs host (i.e. to prevent
167 ipvs routing loops when ipvs is also acting as a real server).
169 nat_icmp_send - BOOLEAN
170 0 - disabled (default)
173 It controls sending icmp error messages (ICMP_DEST_UNREACH)
174 for VS/NAT when the load balancer receives packets from real
175 servers but the connection entries don't exist.
179 not 0 - enabled (default)
181 By default, reject with FRAG_NEEDED all DF packets that exceed
182 the PMTU, irrespective of the forwarding method. For TUN method
183 the flag can be disabled to fragment such packets.
186 0 - disabled (default)
188 The secure_tcp defense is to use a more complicated TCP state
189 transition table. For VS/NAT, it also delays entering the
190 TCP ESTABLISHED state until the three way handshake is completed.
192 The value definition is the same as that of drop_entry and
195 sync_threshold - vector of 2 INTEGERs: sync_threshold, sync_period
198 It sets synchronization threshold, which is the minimum number
199 of incoming packets that a connection needs to receive before
200 the connection will be synchronized. A connection will be
201 synchronized, every time the number of its incoming packets
202 modulus sync_period equals the threshold. The range of the
203 threshold is from 0 to sync_period.
205 When sync_period and sync_refresh_period are 0, send sync only
206 for state changes or only once when pkts matches sync_threshold
208 sync_refresh_period - UNSIGNED INTEGER
211 In seconds, difference in reported connection timer that triggers
212 new sync message. It can be used to avoid sync messages for the
213 specified period (or half of the connection timeout if it is lower)
214 if connection state is not changed since last sync.
216 This is useful for normal connections with high traffic to reduce
217 sync rate. Additionally, retry sync_retries times with period of
218 sync_refresh_period/8.
220 sync_retries - INTEGER
223 Defines sync retries with period of sync_refresh_period/8. Useful
224 to protect against loss of sync messages. The range of the
225 sync_retries is from 0 to 3.
227 sync_qlen_max - UNSIGNED LONG
229 Hard limit for queued sync messages that are not sent yet. It
230 defaults to 1/32 of the memory pages but actually represents
231 number of messages. It will protect us from allocating large
232 parts of memory when the sending rate is lower than the queuing
235 sync_sock_size - INTEGER
238 Configuration of SNDBUF (master) or RCVBUF (slave) socket limit.
239 Default value is 0 (preserve system defaults).
244 The number of threads that master and backup servers can use for
245 sync traffic. Every thread will use single UDP port, thread 0 will
246 use the default port 8848 while last thread will use port
249 snat_reroute - BOOLEAN
251 not 0 - enabled (default)
253 If enabled, recalculate the route of SNATed packets from
254 realservers so that they are routed as if they originate from the
255 director. Otherwise they are routed as if they are forwarded by the
258 If policy routing is in effect then it is possible that the route
259 of a packet originating from a director is routed differently to a
260 packet being forwarded by the director.
262 If policy routing is not in effect then the recalculated route will
263 always be the same as the original route so it is an optimisation
264 to disable snat_reroute and avoid the recalculation.
266 sync_persist_mode - INTEGER
269 Controls the synchronisation of connections when using persistence
271 0: All types of connections are synchronised
272 1: Attempt to reduce the synchronisation traffic depending on
273 the connection type. For persistent services avoid synchronisation
274 for normal connections, do it only for persistence templates.
275 In such case, for TCP and SCTP it may need enabling sloppy_tcp and
276 sloppy_sctp flags on backup servers. For non-persistent services
277 such optimization is not applied, mode 0 is assumed.
279 sync_version - INTEGER
282 The version of the synchronisation protocol used when sending
283 synchronisation messages.
285 0 selects the original synchronisation protocol (version 0). This
286 should be used when sending synchronisation messages to a legacy
287 system that only understands the original synchronisation protocol.
289 1 selects the current synchronisation protocol (version 1). This
290 should be used where possible.
292 Kernels with this sync_version entry are able to receive messages
293 of both version 1 and version 2 of the synchronisation protocol.