| Message ID | 20210819084418.27004-1-dust.li@linux.alibaba.com (mailing list archive) |
|---|---|
| State | Superseded |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net-next,v3] net: ipvs: add sysctl_run_estimation to support disable estimation | expand |
| Context | Check | Description |
|---|---|---|
| netdev/cover_letter | success | Link |
| netdev/fixes_present | success | Link |
| netdev/patch_count | success | Link |
| netdev/tree_selection | success | Clearly marked for net-next |
| netdev/subject_prefix | success | Link |
| netdev/cc_maintainers | warning | 11 maintainers not CCed: davem@davemloft.net dsahern@kernel.org pablo@netfilter.org linux-doc@vger.kernel.org yoshfuji@linux-ipv6.org kuba@kernel.org kadlec@netfilter.org coreteam@netfilter.org fw@strlen.de netfilter-devel@vger.kernel.org corbet@lwn.net |
| netdev/source_inline | success | Was 0 now: 0 |
| netdev/verify_signedoff | success | Link |
| netdev/module_param | success | Was 0 now: 0 |
| netdev/build_32bit | success | Errors and warnings before: 1 this patch: 1 |
| netdev/kdoc | success | Errors and warnings before: 0 this patch: 0 |
| netdev/verify_fixes | success | Link |
| netdev/checkpatch | warning | WARNING: please, no space before tabs |
| netdev/build_allmodconfig_warn | success | Errors and warnings before: 1 this patch: 1 |
| netdev/header_inline | success | Link |
Hello, On Thu, 19 Aug 2021, Dust Li wrote: > estimation_timer will iterater the est_list to do estimation ...will iterate... > for each ipvs stats. When there are lots of services, the > list can be very large. > We observiced estimation_timer() run for more then 200ms on ...We found that estimation_timer() runs for... > a machine with 104 CPU and 50K services. > > yunhong-cgl jiang report the same phenomenon before: > https://www.spinics.net/lists/lvs-devel/msg05426.html > > In some cases(for example a large K8S cluster with many ipvs services), > ipvs estimation may not be needed. So adding a sysctl blob to allow > users to disable this completely. > > Default is: 1 (enable) > > Cc: yunhong-cgl jiang <xintian1976@gmail.com> > Signed-off-by: Dust Li <dust.li@linux.alibaba.com> > --- > v2: Use common sysctl facilities > v3: Fix sysctl_run_estimation() redefine when CONFIG_SYSCTL not enabled > --- Replace above "---" after v3 with empty line, avoid multiple "---" lines. > Documentation/networking/ipvs-sysctl.rst | 17 +++++++++++++++++ > include/net/ip_vs.h | 12 ++++++++++++ > net/netfilter/ipvs/ip_vs_ctl.c | 8 ++++++++ > net/netfilter/ipvs/ip_vs_est.c | 5 +++++ > 4 files changed, 42 insertions(+) > > diff --git a/Documentation/networking/ipvs-sysctl.rst b/Documentation/networking/ipvs-sysctl.rst > index 2afccc63856e..e20f7a27fc85 100644 > --- a/Documentation/networking/ipvs-sysctl.rst > +++ b/Documentation/networking/ipvs-sysctl.rst > @@ -300,3 +300,20 @@ sync_version - INTEGER > > Kernels with this sync_version entry are able to receive messages > of both version 1 and version 2 of the synchronisation protocol. > + > +run_estimation - BOOLEAN > + 0 - disabled > + not 0 - enabled (default) > + > + If disabled, the estimation will be stop, and you can't see > + any update on speed estimation data. > + > + For example > + 'Conns/s Pkts/s Pkts/s Bytes/s Bytes/s' > + those data in /proc/net/ip_vs_stats will always be zero. > + Note, this only affect the speed estimation, the total data > + will still be updated. We can remove the above example > + > + You can always re-enable estimation by setting this value to 1. > + But be carefull, the first estimation after re-enable is not ...careful... > + accurate. > diff --git a/include/net/ip_vs.h b/include/net/ip_vs.h > index 7cb5a1aace40..269f8808f6db 100644 > --- a/include/net/ip_vs.h > +++ b/include/net/ip_vs.h > @@ -931,6 +931,7 @@ struct netns_ipvs { > int sysctl_conn_reuse_mode; > int sysctl_schedule_icmp; > int sysctl_ignore_tunneled; > + int sysctl_run_estimation; scripts/checkpatch.pl --strict /tmp/file.patch reports for extra space just after 'int'. > > /* ip_vs_lblc */ > int sysctl_lblc_expiration; > @@ -1071,6 +1072,11 @@ static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs) > return ipvs->sysctl_cache_bypass; > } > > +static inline int sysctl_run_estimation(struct netns_ipvs *ipvs) > +{ > + return ipvs->sysctl_run_estimation; > +} > + > #else > > static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs) > @@ -1163,6 +1169,11 @@ static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs) > return 0; > } > > +static inline int sysctl_run_estimation(struct netns_ipvs *ipvs) > +{ > + return 1; > +} > + > #endif > > /* IPVS core functions > @@ -1650,6 +1661,7 @@ static inline int ip_vs_confirm_conntrack(struct sk_buff *skb) > static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp) > { > } > + Remove this irrelevant hunk. > #endif /* CONFIG_IP_VS_NFCT */ > > /* Using old conntrack that can not be redirected to another real server? */ > diff --git a/net/netfilter/ipvs/ip_vs_ctl.c b/net/netfilter/ipvs/ip_vs_ctl.c > index c25097092a06..cbea5a68afb5 100644 > --- a/net/netfilter/ipvs/ip_vs_ctl.c > +++ b/net/netfilter/ipvs/ip_vs_ctl.c > @@ -2017,6 +2017,12 @@ static struct ctl_table vs_vars[] = { > .mode = 0644, > .proc_handler = proc_dointvec, > }, > + { > + .procname = "run_estimation", > + .maxlen = sizeof(int), > + .mode = 0644, > + .proc_handler = proc_dointvec, > + }, > #ifdef CONFIG_IP_VS_DEBUG > { > .procname = "debug_level", > @@ -4090,6 +4096,8 @@ static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs) > tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode; > tbl[idx++].data = &ipvs->sysctl_schedule_icmp; > tbl[idx++].data = &ipvs->sysctl_ignore_tunneled; > + ipvs->sysctl_run_estimation = 1; > + tbl[idx++].data = &ipvs->sysctl_run_estimation; > > ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl); > if (ipvs->sysctl_hdr == NULL) { > diff --git a/net/netfilter/ipvs/ip_vs_est.c b/net/netfilter/ipvs/ip_vs_est.c > index 05b8112ffb37..9a1a7af6a186 100644 > --- a/net/netfilter/ipvs/ip_vs_est.c > +++ b/net/netfilter/ipvs/ip_vs_est.c > @@ -100,6 +100,9 @@ static void estimation_timer(struct timer_list *t) > u64 rate; > struct netns_ipvs *ipvs = from_timer(ipvs, t, est_timer); > > + if (!sysctl_run_estimation(ipvs)) > + goto skip; > + > spin_lock(&ipvs->est_lock); > list_for_each_entry(e, &ipvs->est_list, list) { > s = container_of(e, struct ip_vs_stats, est); > @@ -131,6 +134,8 @@ static void estimation_timer(struct timer_list *t) > spin_unlock(&s->lock); > } > spin_unlock(&ipvs->est_lock); > + > +skip: > mod_timer(&ipvs->est_timer, jiffies + 2*HZ); > } > > -- > 2.19.1.3.ge56e4f7 Regards -- Julian Anastasov <ja@ssi.bg>
On Fri, Aug 20, 2021 at 07:53:06AM +0300, Julian Anastasov wrote: > > Hello, > >On Thu, 19 Aug 2021, Dust Li wrote: > >> estimation_timer will iterater the est_list to do estimation > > ...will iterate... > >> for each ipvs stats. When there are lots of services, the >> list can be very large. >> We observiced estimation_timer() run for more then 200ms on > > ...We found that estimation_timer() runs for... OK, thanks for you carefull review, I will fix those details and send a v4. Thanks again ! > >> a machine with 104 CPU and 50K services. >> >> yunhong-cgl jiang report the same phenomenon before: >> https://www.spinics.net/lists/lvs-devel/msg05426.html >> >> In some cases(for example a large K8S cluster with many ipvs services), >> ipvs estimation may not be needed. So adding a sysctl blob to allow >> users to disable this completely. >> >> Default is: 1 (enable) >> >> Cc: yunhong-cgl jiang <xintian1976@gmail.com> >> Signed-off-by: Dust Li <dust.li@linux.alibaba.com> >> --- >> v2: Use common sysctl facilities >> v3: Fix sysctl_run_estimation() redefine when CONFIG_SYSCTL not enabled >> --- > > Replace above "---" after v3 with empty line, >avoid multiple "---" lines. > >> Documentation/networking/ipvs-sysctl.rst | 17 +++++++++++++++++ >> include/net/ip_vs.h | 12 ++++++++++++ >> net/netfilter/ipvs/ip_vs_ctl.c | 8 ++++++++ >> net/netfilter/ipvs/ip_vs_est.c | 5 +++++ >> 4 files changed, 42 insertions(+) >> >> diff --git a/Documentation/networking/ipvs-sysctl.rst b/Documentation/networking/ipvs-sysctl.rst >> index 2afccc63856e..e20f7a27fc85 100644 >> --- a/Documentation/networking/ipvs-sysctl.rst >> +++ b/Documentation/networking/ipvs-sysctl.rst >> @@ -300,3 +300,20 @@ sync_version - INTEGER >> >> Kernels with this sync_version entry are able to receive messages >> of both version 1 and version 2 of the synchronisation protocol. >> + >> +run_estimation - BOOLEAN >> + 0 - disabled >> + not 0 - enabled (default) >> + >> + If disabled, the estimation will be stop, and you can't see >> + any update on speed estimation data. >> + >> + For example >> + 'Conns/s Pkts/s Pkts/s Bytes/s Bytes/s' >> + those data in /proc/net/ip_vs_stats will always be zero. >> + Note, this only affect the speed estimation, the total data >> + will still be updated. > > We can remove the above example > >> + >> + You can always re-enable estimation by setting this value to 1. >> + But be carefull, the first estimation after re-enable is not > > ...careful... > >> + accurate. >> diff --git a/include/net/ip_vs.h b/include/net/ip_vs.h >> index 7cb5a1aace40..269f8808f6db 100644 >> --- a/include/net/ip_vs.h >> +++ b/include/net/ip_vs.h >> @@ -931,6 +931,7 @@ struct netns_ipvs { >> int sysctl_conn_reuse_mode; >> int sysctl_schedule_icmp; >> int sysctl_ignore_tunneled; >> + int sysctl_run_estimation; > > scripts/checkpatch.pl --strict /tmp/file.patch >reports for extra space just after 'int'. > >> >> /* ip_vs_lblc */ >> int sysctl_lblc_expiration; >> @@ -1071,6 +1072,11 @@ static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs) >> return ipvs->sysctl_cache_bypass; >> } >> >> +static inline int sysctl_run_estimation(struct netns_ipvs *ipvs) >> +{ >> + return ipvs->sysctl_run_estimation; >> +} >> + >> #else >> >> static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs) >> @@ -1163,6 +1169,11 @@ static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs) >> return 0; >> } >> >> +static inline int sysctl_run_estimation(struct netns_ipvs *ipvs) >> +{ >> + return 1; >> +} >> + >> #endif >> >> /* IPVS core functions >> @@ -1650,6 +1661,7 @@ static inline int ip_vs_confirm_conntrack(struct sk_buff *skb) >> static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp) >> { >> } >> + > > Remove this irrelevant hunk. > >> #endif /* CONFIG_IP_VS_NFCT */ >> >> /* Using old conntrack that can not be redirected to another real server? */ >> diff --git a/net/netfilter/ipvs/ip_vs_ctl.c b/net/netfilter/ipvs/ip_vs_ctl.c >> index c25097092a06..cbea5a68afb5 100644 >> --- a/net/netfilter/ipvs/ip_vs_ctl.c >> +++ b/net/netfilter/ipvs/ip_vs_ctl.c >> @@ -2017,6 +2017,12 @@ static struct ctl_table vs_vars[] = { >> .mode = 0644, >> .proc_handler = proc_dointvec, >> }, >> + { >> + .procname = "run_estimation", >> + .maxlen = sizeof(int), >> + .mode = 0644, >> + .proc_handler = proc_dointvec, >> + }, >> #ifdef CONFIG_IP_VS_DEBUG >> { >> .procname = "debug_level", >> @@ -4090,6 +4096,8 @@ static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs) >> tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode; >> tbl[idx++].data = &ipvs->sysctl_schedule_icmp; >> tbl[idx++].data = &ipvs->sysctl_ignore_tunneled; >> + ipvs->sysctl_run_estimation = 1; >> + tbl[idx++].data = &ipvs->sysctl_run_estimation; >> >> ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl); >> if (ipvs->sysctl_hdr == NULL) { >> diff --git a/net/netfilter/ipvs/ip_vs_est.c b/net/netfilter/ipvs/ip_vs_est.c >> index 05b8112ffb37..9a1a7af6a186 100644 >> --- a/net/netfilter/ipvs/ip_vs_est.c >> +++ b/net/netfilter/ipvs/ip_vs_est.c >> @@ -100,6 +100,9 @@ static void estimation_timer(struct timer_list *t) >> u64 rate; >> struct netns_ipvs *ipvs = from_timer(ipvs, t, est_timer); >> >> + if (!sysctl_run_estimation(ipvs)) >> + goto skip; >> + >> spin_lock(&ipvs->est_lock); >> list_for_each_entry(e, &ipvs->est_list, list) { >> s = container_of(e, struct ip_vs_stats, est); >> @@ -131,6 +134,8 @@ static void estimation_timer(struct timer_list *t) >> spin_unlock(&s->lock); >> } >> spin_unlock(&ipvs->est_lock); >> + >> +skip: >> mod_timer(&ipvs->est_timer, jiffies + 2*HZ); >> } >> >> -- >> 2.19.1.3.ge56e4f7 > >Regards > >-- >Julian Anastasov <ja@ssi.bg>
diff --git a/Documentation/networking/ipvs-sysctl.rst b/Documentation/networking/ipvs-sysctl.rst index 2afccc63856e..e20f7a27fc85 100644 --- a/Documentation/networking/ipvs-sysctl.rst +++ b/Documentation/networking/ipvs-sysctl.rst @@ -300,3 +300,20 @@ sync_version - INTEGER Kernels with this sync_version entry are able to receive messages of both version 1 and version 2 of the synchronisation protocol. + +run_estimation - BOOLEAN + 0 - disabled + not 0 - enabled (default) + + If disabled, the estimation will be stop, and you can't see + any update on speed estimation data. + + For example + 'Conns/s Pkts/s Pkts/s Bytes/s Bytes/s' + those data in /proc/net/ip_vs_stats will always be zero. + Note, this only affect the speed estimation, the total data + will still be updated. + + You can always re-enable estimation by setting this value to 1. + But be carefull, the first estimation after re-enable is not + accurate. diff --git a/include/net/ip_vs.h b/include/net/ip_vs.h index 7cb5a1aace40..269f8808f6db 100644 --- a/include/net/ip_vs.h +++ b/include/net/ip_vs.h @@ -931,6 +931,7 @@ struct netns_ipvs { int sysctl_conn_reuse_mode; int sysctl_schedule_icmp; int sysctl_ignore_tunneled; + int sysctl_run_estimation; /* ip_vs_lblc */ int sysctl_lblc_expiration; @@ -1071,6 +1072,11 @@ static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs) return ipvs->sysctl_cache_bypass; } +static inline int sysctl_run_estimation(struct netns_ipvs *ipvs) +{ + return ipvs->sysctl_run_estimation; +} + #else static inline int sysctl_sync_threshold(struct netns_ipvs *ipvs) @@ -1163,6 +1169,11 @@ static inline int sysctl_cache_bypass(struct netns_ipvs *ipvs) return 0; } +static inline int sysctl_run_estimation(struct netns_ipvs *ipvs) +{ + return 1; +} + #endif /* IPVS core functions @@ -1650,6 +1661,7 @@ static inline int ip_vs_confirm_conntrack(struct sk_buff *skb) static inline void ip_vs_conn_drop_conntrack(struct ip_vs_conn *cp) { } + #endif /* CONFIG_IP_VS_NFCT */ /* Using old conntrack that can not be redirected to another real server? */ diff --git a/net/netfilter/ipvs/ip_vs_ctl.c b/net/netfilter/ipvs/ip_vs_ctl.c index c25097092a06..cbea5a68afb5 100644 --- a/net/netfilter/ipvs/ip_vs_ctl.c +++ b/net/netfilter/ipvs/ip_vs_ctl.c @@ -2017,6 +2017,12 @@ static struct ctl_table vs_vars[] = { .mode = 0644, .proc_handler = proc_dointvec, }, + { + .procname = "run_estimation", + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = proc_dointvec, + }, #ifdef CONFIG_IP_VS_DEBUG { .procname = "debug_level", @@ -4090,6 +4096,8 @@ static int __net_init ip_vs_control_net_init_sysctl(struct netns_ipvs *ipvs) tbl[idx++].data = &ipvs->sysctl_conn_reuse_mode; tbl[idx++].data = &ipvs->sysctl_schedule_icmp; tbl[idx++].data = &ipvs->sysctl_ignore_tunneled; + ipvs->sysctl_run_estimation = 1; + tbl[idx++].data = &ipvs->sysctl_run_estimation; ipvs->sysctl_hdr = register_net_sysctl(net, "net/ipv4/vs", tbl); if (ipvs->sysctl_hdr == NULL) { diff --git a/net/netfilter/ipvs/ip_vs_est.c b/net/netfilter/ipvs/ip_vs_est.c index 05b8112ffb37..9a1a7af6a186 100644 --- a/net/netfilter/ipvs/ip_vs_est.c +++ b/net/netfilter/ipvs/ip_vs_est.c @@ -100,6 +100,9 @@ static void estimation_timer(struct timer_list *t) u64 rate; struct netns_ipvs *ipvs = from_timer(ipvs, t, est_timer); + if (!sysctl_run_estimation(ipvs)) + goto skip; + spin_lock(&ipvs->est_lock); list_for_each_entry(e, &ipvs->est_list, list) { s = container_of(e, struct ip_vs_stats, est); @@ -131,6 +134,8 @@ static void estimation_timer(struct timer_list *t) spin_unlock(&s->lock); } spin_unlock(&ipvs->est_lock); + +skip: mod_timer(&ipvs->est_timer, jiffies + 2*HZ); }
estimation_timer will iterater the est_list to do estimation for each ipvs stats. When there are lots of services, the list can be very large. We observiced estimation_timer() run for more then 200ms on a machine with 104 CPU and 50K services. yunhong-cgl jiang report the same phenomenon before: https://www.spinics.net/lists/lvs-devel/msg05426.html In some cases(for example a large K8S cluster with many ipvs services), ipvs estimation may not be needed. So adding a sysctl blob to allow users to disable this completely. Default is: 1 (enable) Cc: yunhong-cgl jiang <xintian1976@gmail.com> Signed-off-by: Dust Li <dust.li@linux.alibaba.com> --- v2: Use common sysctl facilities v3: Fix sysctl_run_estimation() redefine when CONFIG_SYSCTL not enabled --- Documentation/networking/ipvs-sysctl.rst | 17 +++++++++++++++++ include/net/ip_vs.h | 12 ++++++++++++ net/netfilter/ipvs/ip_vs_ctl.c | 8 ++++++++ net/netfilter/ipvs/ip_vs_est.c | 5 +++++ 4 files changed, 42 insertions(+)