333 lines · plain
1.. SPDX-License-Identifier: GPL-2.02 3===========4IPvs-sysctl5===========6 7/proc/sys/net/ipv4/vs/* Variables:8==================================9 10am_droprate - INTEGER11 default 1012 13 It sets the always mode drop rate, which is used in the mode 314 of the drop_rate defense.15 16amemthresh - INTEGER17 default 102418 19 It sets the available memory threshold (in pages), which is20 used in the automatic modes of defense. When there is no21 enough available memory, the respective strategy will be22 enabled and the variable is automatically set to 2, otherwise23 the strategy is disabled and the variable is set to 1.24 25backup_only - BOOLEAN26 - 0 - disabled (default)27 - not 0 - enabled28 29 If set, disable the director function while the server is30 in backup mode to avoid packet loops for DR/TUN methods.31 32conn_reuse_mode - INTEGER33 1 - default34 35 Controls how ipvs will deal with connections that are detected36 port reuse. It is a bitmap, with the values being:37 38 0: disable any special handling on port reuse. The new39 connection will be delivered to the same real server that was40 servicing the previous connection.41 42 bit 1: enable rescheduling of new connections when it is safe.43 That is, whenever expire_nodest_conn and for TCP sockets, when44 the connection is in TIME_WAIT state (which is only possible if45 you use NAT mode).46 47 bit 2: it is bit 1 plus, for TCP connections, when connections48 are in FIN_WAIT state, as this is the last state seen by load49 balancer in Direct Routing mode. This bit helps on adding new50 real servers to a very busy cluster.51 52conntrack - BOOLEAN53 - 0 - disabled (default)54 - not 0 - enabled55 56 If set, maintain connection tracking entries for57 connections handled by IPVS.58 59 This should be enabled if connections handled by IPVS are to be60 also handled by stateful firewall rules. That is, iptables rules61 that make use of connection tracking. It is a performance62 optimisation to disable this setting otherwise.63 64 Connections handled by the IPVS FTP application module65 will have connection tracking entries regardless of this setting.66 67 Only available when IPVS is compiled with CONFIG_IP_VS_NFCT enabled.68 69cache_bypass - BOOLEAN70 - 0 - disabled (default)71 - not 0 - enabled72 73 If it is enabled, forward packets to the original destination74 directly when no cache server is available and destination75 address is not local (iph->daddr is RTN_UNICAST). It is mostly76 used in transparent web cache cluster.77 78debug_level - INTEGER79 - 0 - transmission error messages (default)80 - 1 - non-fatal error messages81 - 2 - configuration82 - 3 - destination trash83 - 4 - drop entry84 - 5 - service lookup85 - 6 - scheduling86 - 7 - connection new/expire, lookup and synchronization87 - 8 - state transition88 - 9 - binding destination, template checks and applications89 - 10 - IPVS packet transmission90 - 11 - IPVS packet handling (ip_vs_in/ip_vs_out)91 - 12 or more - packet traversal92 93 Only available when IPVS is compiled with CONFIG_IP_VS_DEBUG enabled.94 95 Higher debugging levels include the messages for lower debugging96 levels, so setting debug level 2, includes level 0, 1 and 297 messages. Thus, logging becomes more and more verbose the higher98 the level.99 100drop_entry - INTEGER101 - 0 - disabled (default)102 103 The drop_entry defense is to randomly drop entries in the104 connection hash table, just in order to collect back some105 memory for new connections. In the current code, the106 drop_entry procedure can be activated every second, then it107 randomly scans 1/32 of the whole and drops entries that are in108 the SYN-RECV/SYNACK state, which should be effective against109 syn-flooding attack.110 111 The valid values of drop_entry are from 0 to 3, where 0 means112 that this strategy is always disabled, 1 and 2 mean automatic113 modes (when there is no enough available memory, the strategy114 is enabled and the variable is automatically set to 2,115 otherwise the strategy is disabled and the variable is set to116 1), and 3 means that the strategy is always enabled.117 118drop_packet - INTEGER119 - 0 - disabled (default)120 121 The drop_packet defense is designed to drop 1/rate packets122 before forwarding them to real servers. If the rate is 1, then123 drop all the incoming packets.124 125 The value definition is the same as that of the drop_entry. In126 the automatic mode, the rate is determined by the follow127 formula: rate = amemthresh / (amemthresh - available_memory)128 when available memory is less than the available memory129 threshold. When the mode 3 is set, the always mode drop rate130 is controlled by the /proc/sys/net/ipv4/vs/am_droprate.131 132est_cpulist - CPULIST133 Allowed CPUs for estimation kthreads134 135 Syntax: standard cpulist format136 empty list - stop kthread tasks and estimation137 default - the system's housekeeping CPUs for kthreads138 139 Example:140 "all": all possible CPUs141 "0-N": all possible CPUs, N denotes last CPU number142 "0,1-N:1/2": first and all CPUs with odd number143 "": empty list144 145est_nice - INTEGER146 default 0147 Valid range: -20 (more favorable) .. 19 (less favorable)148 149 Niceness value to use for the estimation kthreads (scheduling150 priority)151 152expire_nodest_conn - BOOLEAN153 - 0 - disabled (default)154 - not 0 - enabled155 156 The default value is 0, the load balancer will silently drop157 packets when its destination server is not available. It may158 be useful, when user-space monitoring program deletes the159 destination server (because of server overload or wrong160 detection) and add back the server later, and the connections161 to the server can continue.162 163 If this feature is enabled, the load balancer will expire the164 connection immediately when a packet arrives and its165 destination server is not available, then the client program166 will be notified that the connection is closed. This is167 equivalent to the feature some people requires to flush168 connections when its destination is not available.169 170expire_quiescent_template - BOOLEAN171 - 0 - disabled (default)172 - not 0 - enabled173 174 When set to a non-zero value, the load balancer will expire175 persistent templates when the destination server is quiescent.176 This may be useful, when a user makes a destination server177 quiescent by setting its weight to 0 and it is desired that178 subsequent otherwise persistent connections are sent to a179 different destination server. By default new persistent180 connections are allowed to quiescent destination servers.181 182 If this feature is enabled, the load balancer will expire the183 persistence template if it is to be used to schedule a new184 connection and the destination server is quiescent.185 186ignore_tunneled - BOOLEAN187 - 0 - disabled (default)188 - not 0 - enabled189 190 If set, ipvs will set the ipvs_property on all packets which are of191 unrecognized protocols. This prevents us from routing tunneled192 protocols like ipip, which is useful to prevent rescheduling193 packets that have been tunneled to the ipvs host (i.e. to prevent194 ipvs routing loops when ipvs is also acting as a real server).195 196nat_icmp_send - BOOLEAN197 - 0 - disabled (default)198 - not 0 - enabled199 200 It controls sending icmp error messages (ICMP_DEST_UNREACH)201 for VS/NAT when the load balancer receives packets from real202 servers but the connection entries don't exist.203 204pmtu_disc - BOOLEAN205 - 0 - disabled206 - not 0 - enabled (default)207 208 By default, reject with FRAG_NEEDED all DF packets that exceed209 the PMTU, irrespective of the forwarding method. For TUN method210 the flag can be disabled to fragment such packets.211 212secure_tcp - INTEGER213 - 0 - disabled (default)214 215 The secure_tcp defense is to use a more complicated TCP state216 transition table. For VS/NAT, it also delays entering the217 TCP ESTABLISHED state until the three way handshake is completed.218 219 The value definition is the same as that of drop_entry and220 drop_packet.221 222sync_threshold - vector of 2 INTEGERs: sync_threshold, sync_period223 default 3 50224 225 It sets synchronization threshold, which is the minimum number226 of incoming packets that a connection needs to receive before227 the connection will be synchronized. A connection will be228 synchronized, every time the number of its incoming packets229 modulus sync_period equals the threshold. The range of the230 threshold is from 0 to sync_period.231 232 When sync_period and sync_refresh_period are 0, send sync only233 for state changes or only once when pkts matches sync_threshold234 235sync_refresh_period - UNSIGNED INTEGER236 default 0237 238 In seconds, difference in reported connection timer that triggers239 new sync message. It can be used to avoid sync messages for the240 specified period (or half of the connection timeout if it is lower)241 if connection state is not changed since last sync.242 243 This is useful for normal connections with high traffic to reduce244 sync rate. Additionally, retry sync_retries times with period of245 sync_refresh_period/8.246 247sync_retries - INTEGER248 default 0249 250 Defines sync retries with period of sync_refresh_period/8. Useful251 to protect against loss of sync messages. The range of the252 sync_retries is from 0 to 3.253 254sync_qlen_max - UNSIGNED LONG255 256 Hard limit for queued sync messages that are not sent yet. It257 defaults to 1/32 of the memory pages but actually represents258 number of messages. It will protect us from allocating large259 parts of memory when the sending rate is lower than the queuing260 rate.261 262sync_sock_size - INTEGER263 default 0264 265 Configuration of SNDBUF (master) or RCVBUF (slave) socket limit.266 Default value is 0 (preserve system defaults).267 268sync_ports - INTEGER269 default 1270 271 The number of threads that master and backup servers can use for272 sync traffic. Every thread will use single UDP port, thread 0 will273 use the default port 8848 while last thread will use port274 8848+sync_ports-1.275 276snat_reroute - BOOLEAN277 - 0 - disabled278 - not 0 - enabled (default)279 280 If enabled, recalculate the route of SNATed packets from281 realservers so that they are routed as if they originate from the282 director. Otherwise they are routed as if they are forwarded by the283 director.284 285 If policy routing is in effect then it is possible that the route286 of a packet originating from a director is routed differently to a287 packet being forwarded by the director.288 289 If policy routing is not in effect then the recalculated route will290 always be the same as the original route so it is an optimisation291 to disable snat_reroute and avoid the recalculation.292 293sync_persist_mode - INTEGER294 default 0295 296 Controls the synchronisation of connections when using persistence297 298 0: All types of connections are synchronised299 300 1: Attempt to reduce the synchronisation traffic depending on301 the connection type. For persistent services avoid synchronisation302 for normal connections, do it only for persistence templates.303 In such case, for TCP and SCTP it may need enabling sloppy_tcp and304 sloppy_sctp flags on backup servers. For non-persistent services305 such optimization is not applied, mode 0 is assumed.306 307sync_version - INTEGER308 default 1309 310 The version of the synchronisation protocol used when sending311 synchronisation messages.312 313 0 selects the original synchronisation protocol (version 0). This314 should be used when sending synchronisation messages to a legacy315 system that only understands the original synchronisation protocol.316 317 1 selects the current synchronisation protocol (version 1). This318 should be used where possible.319 320 Kernels with this sync_version entry are able to receive messages321 of both version 1 and version 2 of the synchronisation protocol.322 323run_estimation - BOOLEAN324 0 - disabled325 not 0 - enabled (default)326 327 If disabled, the estimation will be suspended and kthread tasks328 stopped.329 330 You can always re-enable estimation by setting this value to 1.331 But be careful, the first estimation after re-enable is not332 accurate.333