200ms delays with SCTP streaming data
Hi, it's me again with another strange issue. In case you didn't figure it out before, I'm working on a library that supports all different types of stream I/O, and SCTP is one supported building block. I noticed when I stacked a multiplexer layer on top of SCTP I started getting timeouts occasionally. It took a bit, but I finally realized that I was getting 200ms delays occasionally between sending a packet and receiving a packet. I verified this with a trace right at the sctp_send() and sctp_recvmsg() calls. It doesn't seem to be regular in any way I can see, but it happens often enough to cause issues.
If I replace the SCTP block with a TCP block, it works fine, and pretty much all the code is the same except where it does the read and write calls (including the epoll() usage, and I have also switched to select() and it has the same issue). The write calls don't seem to be the issue, I see two back-to-back writes a few microseconds apart and see a 200ms delay between the messages on the receive side.
The test in question sets up two connections and does a big simultaneous bidirectional transfer. The test app has 4 threads waiting on epoll() handling data and writing data.
And the delay is always ~200ms. Which sounds suspicious.
It's not using sctp_sendv() at the moment, as the systems I'm running on don't have that yet. But the library does have support if it sees it is available.
So I don't think it's my library; I've stared at it a bunch (and found a few other bugs) but I can't reconcile this one. There are no timers that would cause this in the code in question. Just basically an epoll() call waiting on data and receive processing that is comparing data, along with write processing that is sending the same data.
Anyway, I haven't tried to create a small reproducer; I thought I would report it first and see if anything rang a bell. I tried this on a recent kernel and got the same issue.
The library is at https://github.com/cminyard/gensio. I'd need to provide a patch for the tracing.
-corey
Marcelo Leitner
maintainerQuoted 17 lines · Corey Minyard
Hi, it's me again with another strange issue. In case you didn't figure it out before, I'm working on a library that supports all different types of stream I/O, and SCTP is one supported building block. I noticed when I stacked a multiplexer layer on top of SCTP I started getting timeouts occasionally. It took a bit, but I finally realized that I was getting 200ms delays occasionally between sending a packet and receiving a packet. I verified this with a trace right at the
sctp_send()andsctp_recvmsg()calls. It doesn't seem to be regular in any way I can see, but it happens often enough to cause issues.If I replace the SCTP block with a TCP block, it works fine, and pretty much all the code is the same except where it does the read and write calls (including the
epoll()usage, and I have also switched toselect()and it has the same issue). The write calls don't seem to be the issue, I see two back-to-back writes a few microseconds apart and see a 200ms delay between the messages on the receive side.
Quoted 5 lines
The test in question sets up two connections and does a big simultaneous bidirectional transfer. The test app has 4 threads waiting on
epoll()handling data and writing data.And the delay is always ~200ms. Which sounds suspicious.
That can be the delayed sack timer, in kernel.
/* Delayed sack timer - 200ms */
#define SCTP_DEFAULT_TIMEOUT_SACK (200)
You may tweak the sysctl net.sctp.sack_timeout and see if changes accordingly, or via SCTP_PEER_ADDR_PARAMS or even enable immediate ack (by setting SPP_SACKDELAY_DISABLE)
Quoted 7 lines
It's not using
sctp_sendv()at the moment, as the systems I'm running on don't have that yet. But the library does have support if it sees it is available.So I don't think it's my library; I've stared at it a bunch (and found a few other bugs) but I can't reconcile this one. There are no timers
Nice.
Quoted 7 lines
that would cause this in the code in question. Just basically an
epoll()call waiting on data and receive processing that is comparing data, along with write processing that is sending the same data.Anyway, I haven't tried to create a small reproducer; I thought I would report it first and see if anything rang a bell. I tried this on a recent kernel and got the same issue.
I guess a combination of xmit rate, msg and buffer sizes and packet drops can lead to this delay. I've seen it happening, but didn't have the time to track it down back then.
That said, remember that Linux SCTP doesn't support buffer auto-tuning. So considering you're running a stress test, you probably want to adjust them accordingly manually to avoid packet drops.
Marcelo
Quoted 5 lines
The library is at https://github.com/cminyard/gensio. I'd need to provide a patch for the tracing.
-corey
Quoted 24 lines · Marcelo Leitner
Quoted 17 lines · Corey Minyard
Hi, it's me again with another strange issue. In case you didn't figure it out before, I'm working on a library that supports all different types of stream I/O, and SCTP is one supported building block. I noticed when I stacked a multiplexer layer on top of SCTP I started getting timeouts occasionally. It took a bit, but I finally realized that I was getting 200ms delays occasionally between sending a packet and receiving a packet. I verified this with a trace right at the
sctp_send()andsctp_recvmsg()calls. It doesn't seem to be regular in any way I can see, but it happens often enough to cause issues.If I replace the SCTP block with a TCP block, it works fine, and pretty much all the code is the same except where it does the read and write calls (including the
epoll()usage, and I have also switched toselect()and it has the same issue). The write calls don't seem to be the issue, I see two back-to-back writes a few microseconds apart and see a 200ms delay between the messages on the receive side.Quoted 5 lines
The test in question sets up two connections and does a big simultaneous bidirectional transfer. The test app has 4 threads waiting on
epoll()handling data and writing data.And the delay is always ~200ms. Which sounds suspicious.
Quoted 7 lines
That can be the delayed sack timer, in kernel.
/* Delayed sack timer - 200ms */
#defineSCTP_DEFAULT_TIMEOUT_SACK(200)You may tweak the sysctl net.sctp.sack_timeout and see if changes accordingly, or via
SCTP_PEER_ADDR_PARAMSor even enable immediate ack (by settingSPP_SACKDELAY_DISABLE)
Ok, setting SPP_SACKDELAY_DISABLE does make the problem go away.
Quoted 15 lines
Quoted 8 lines
It's not using
sctp_sendv()at the moment, as the systems I'm running on don't have that yet. But the library does have support if it sees it is available.So I don't think it's my library; I've stared at it a bunch (and found a few other bugs) but I can't reconcile this one. There are no timers
Nice.
Quoted 4 lines
that would cause this in the code in question. Just basically an
epoll()call waiting on data and receive processing that is comparing data, along with write processing that is sending the same data.
Quoted 7 lines
Quoted 3 lines
Anyway, I haven't tried to create a small reproducer; I thought I would report it first and see if anything rang a bell. I tried this on a recent kernel and got the same issue.
I guess a combination of xmit rate, msg and buffer sizes and packet drops can lead to this delay. I've seen it happening, but didn't have the time to track it down back then.
There should be no packet drops. It's running across localhost, and flow control in the multiplex layer as it's set up for the tests limits the maximum outstanding data to 1024 bytes. With overhead and flow control messages it's maybe 1050 bytes of data that would ever be unacked. (It's not really testing throughput, it's testing the inner workings of the multiplexing protocol.)
If I understand this correctly per the RFCs, by default if there are two messages outstanding, it will send an sack immediately. Otherwise it waits 200ms. I'm guessing what is happening is that that SCTP sends a sack and then receives one more message and the upper layer stops because of flow control. Then the sack comes out in 200ms and things continue.
So I think I can figure out how to make this work smoothly. I assume this doesn't happen in TCP because all packets carry an ack, and there is data flowing both ways all the time.
Thanks,
-corey
Quoted 12 lines
That said, remember that Linux SCTP doesn't support buffer auto-tuning. So considering you're running a stress test, you probably want to adjust them accordingly manually to avoid packet drops.
Marcelo
Quoted 5 lines
The library is at https://github.com/cminyard/gensio. I'd need to provide a patch for the tracing.
-corey
Quoted 65 lines · Corey Minyard
Quoted 24 lines · Marcelo Leitner
Quoted 17 lines · Corey Minyard
Hi, it's me again with another strange issue. In case you didn't figure it out before, I'm working on a library that supports all different types of stream I/O, and SCTP is one supported building block. I noticed when I stacked a multiplexer layer on top of SCTP I started getting timeouts occasionally. It took a bit, but I finally realized that I was getting 200ms delays occasionally between sending a packet and receiving a packet. I verified this with a trace right at the
sctp_send()andsctp_recvmsg()calls. It doesn't seem to be regular in any way I can see, but it happens often enough to cause issues.If I replace the SCTP block with a TCP block, it works fine, and pretty much all the code is the same except where it does the read and write calls (including the
epoll()usage, and I have also switched toselect()and it has the same issue). The write calls don't seem to be the issue, I see two back-to-back writes a few microseconds apart and see a 200ms delay between the messages on the receive side.Quoted 5 lines
The test in question sets up two connections and does a big simultaneous bidirectional transfer. The test app has 4 threads waiting on
epoll()handling data and writing data.And the delay is always ~200ms. Which sounds suspicious.
Quoted 7 lines
That can be the delayed sack timer, in kernel.
/* Delayed sack timer - 200ms */
#defineSCTP_DEFAULT_TIMEOUT_SACK(200)You may tweak the sysctl net.sctp.sack_timeout and see if changes accordingly, or via
SCTP_PEER_ADDR_PARAMSor even enable immediate ack (by settingSPP_SACKDELAY_DISABLE)Ok, setting
SPP_SACKDELAY_DISABLEdoes make the problem go away.Quoted 15 lines
Quoted 8 lines
It's not using
sctp_sendv()at the moment, as the systems I'm running on don't have that yet. But the library does have support if it sees it is available.So I don't think it's my library; I've stared at it a bunch (and found a few other bugs) but I can't reconcile this one. There are no timers
Nice.
Quoted 4 lines
that would cause this in the code in question. Just basically an
epoll()call waiting on data and receive processing that is comparing data, along with write processing that is sending the same data.Quoted 7 lines
Quoted 3 lines
Anyway, I haven't tried to create a small reproducer; I thought I would report it first and see if anything rang a bell. I tried this on a recent kernel and got the same issue.
I guess a combination of xmit rate, msg and buffer sizes and packet drops can lead to this delay. I've seen it happening, but didn't have the time to track it down back then.
There should be no packet drops. It's running across localhost, and flow control in the multiplex layer as it's set up for the tests limits the maximum outstanding data to 1024 bytes. With overhead and flow control messages it's maybe 1050 bytes of data that would ever be unacked. (It's not really testing throughput, it's testing the inner workings of the multiplexing protocol.)
Quoted 6 lines
If I understand this correctly per the RFCs, by default if there are two messages outstanding, it will send an sack immediately. Otherwise it waits 200ms. I'm guessing what is happening is that that SCTP sends a sack and then receives one more message and the upper layer stops because of flow control. Then the sack comes out in 200ms and things continue.
Actually, that still doesn't make sense. The lack of a sack shouldn't keep anything from sending unless the congestion window is closed, which shouldn't happen in this case. Am I correct?
-corey
Quoted 21 lines
So I think I can figure out how to make this work smoothly. I assume this doesn't happen in TCP because all packets carry an ack, and there is data flowing both ways all the time.
Thanks,
-corey
Quoted 12 lines
That said, remember that Linux SCTP doesn't support buffer auto-tuning. So considering you're running a stress test, you probably want to adjust them accordingly manually to avoid packet drops.
Marcelo
Quoted 5 lines
The library is at https://github.com/cminyard/gensio. I'd need to provide a patch for the tracing.
-corey
Quoted 75 lines · Corey Minyard
Quoted 65 lines · Corey Minyard
Quoted 24 lines · Marcelo Leitner
Quoted 17 lines · Corey Minyard
Hi, it's me again with another strange issue. In case you didn't figure it out before, I'm working on a library that supports all different types of stream I/O, and SCTP is one supported building block. I noticed when I stacked a multiplexer layer on top of SCTP I started getting timeouts occasionally. It took a bit, but I finally realized that I was getting 200ms delays occasionally between sending a packet and receiving a packet. I verified this with a trace right at the
sctp_send()andsctp_recvmsg()calls. It doesn't seem to be regular in any way I can see, but it happens often enough to cause issues.If I replace the SCTP block with a TCP block, it works fine, and pretty much all the code is the same except where it does the read and write calls (including the
epoll()usage, and I have also switched toselect()and it has the same issue). The write calls don't seem to be the issue, I see two back-to-back writes a few microseconds apart and see a 200ms delay between the messages on the receive side.Quoted 5 lines
The test in question sets up two connections and does a big simultaneous bidirectional transfer. The test app has 4 threads waiting on
epoll()handling data and writing data.And the delay is always ~200ms. Which sounds suspicious.
Quoted 7 lines
That can be the delayed sack timer, in kernel.
/* Delayed sack timer - 200ms */
#defineSCTP_DEFAULT_TIMEOUT_SACK(200)You may tweak the sysctl net.sctp.sack_timeout and see if changes accordingly, or via
SCTP_PEER_ADDR_PARAMSor even enable immediate ack (by settingSPP_SACKDELAY_DISABLE)Ok, setting
SPP_SACKDELAY_DISABLEdoes make the problem go away.Quoted 15 lines
Quoted 8 lines
It's not using
sctp_sendv()at the moment, as the systems I'm running on don't have that yet. But the library does have support if it sees it is available.So I don't think it's my library; I've stared at it a bunch (and found a few other bugs) but I can't reconcile this one. There are no timers
Nice.
Quoted 4 lines
that would cause this in the code in question. Just basically an
epoll()call waiting on data and receive processing that is comparing data, along with write processing that is sending the same data.Quoted 7 lines
Quoted 3 lines
Anyway, I haven't tried to create a small reproducer; I thought I would report it first and see if anything rang a bell. I tried this on a recent kernel and got the same issue.
I guess a combination of xmit rate, msg and buffer sizes and packet drops can lead to this delay. I've seen it happening, but didn't have the time to track it down back then.
There should be no packet drops. It's running across localhost, and flow control in the multiplex layer as it's set up for the tests limits the maximum outstanding data to 1024 bytes. With overhead and flow control messages it's maybe 1050 bytes of data that would ever be unacked. (It's not really testing throughput, it's testing the inner workings of the multiplexing protocol.)
Quoted 6 lines
If I understand this correctly per the RFCs, by default if there are two messages outstanding, it will send an sack immediately. Otherwise it waits 200ms. I'm guessing what is happening is that that SCTP sends a sack and then receives one more message and the upper layer stops because of flow control. Then the sack comes out in 200ms and things continue.
Quoted 3 lines
Actually, that still doesn't make sense. The lack of a sack shouldn't keep anything from sending unless the congestion window is closed, which shouldn't happen in this case. Am I correct?
I guess you still have the Nagle algorithm enabled. Try enabled the SCTP_NODELAY socket option: https://tools.ietf.org/html/rfc6458#section-8.1.5 at the sender side.
It is enabled by default and will delay the sending of packet if they are not large enough (an implementation decision) and there is outstanding data.
Best regards
Michael
Quoted 24 lines
-corey
Quoted 21 lines
So I think I can figure out how to make this work smoothly. I assume this doesn't happen in TCP because all packets carry an ack, and there is data flowing both ways all the time.
Thanks,
-corey
Quoted 12 lines
That said, remember that Linux SCTP doesn't support buffer auto-tuning. So considering you're running a stress test, you probably want to adjust them accordingly manually to avoid packet drops.
Marcelo
Quoted 5 lines
The library is at https://github.com/cminyard/gensio. I'd need to provide a patch for the tracing.
-corey
Quoted 75 lines · Michael Tuexen
Quoted 75 lines · Corey Minyard
Quoted 71 lines · Corey Minyard
Quoted 24 lines · Marcelo Leitner
Quoted 17 lines · Corey Minyard
Hi, it's me again with another strange issue. In case you didn't figure it out before, I'm working on a library that supports all different types of stream I/O, and SCTP is one supported building block. I noticed when I stacked a multiplexer layer on top of SCTP I started getting timeouts occasionally. It took a bit, but I finally realized that I was getting 200ms delays occasionally between sending a packet and receiving a packet. I verified this with a trace right at the
sctp_send()andsctp_recvmsg()calls. It doesn't seem to be regular in any way I can see, but it happens often enough to cause issues.If I replace the SCTP block with a TCP block, it works fine, and pretty much all the code is the same except where it does the read and write calls (including the
epoll()usage, and I have also switched toselect()and it has the same issue). The write calls don't seem to be the issue, I see two back-to-back writes a few microseconds apart and see a 200ms delay between the messages on the receive side.Quoted 5 lines
The test in question sets up two connections and does a big simultaneous bidirectional transfer. The test app has 4 threads waiting on
epoll()handling data and writing data.And the delay is always ~200ms. Which sounds suspicious.
Quoted 7 lines
That can be the delayed sack timer, in kernel.
/* Delayed sack timer - 200ms */
#defineSCTP_DEFAULT_TIMEOUT_SACK(200)You may tweak the sysctl net.sctp.sack_timeout and see if changes accordingly, or via
SCTP_PEER_ADDR_PARAMSor even enable immediate ack (by settingSPP_SACKDELAY_DISABLE)Ok, setting
SPP_SACKDELAY_DISABLEdoes make the problem go away.Quoted 15 lines
Quoted 8 lines
It's not using
sctp_sendv()at the moment, as the systems I'm running on don't have that yet. But the library does have support if it sees it is available.So I don't think it's my library; I've stared at it a bunch (and found a few other bugs) but I can't reconcile this one. There are no timers
Nice.
Quoted 4 lines
that would cause this in the code in question. Just basically an
epoll()call waiting on data and receive processing that is comparing data, along with write processing that is sending the same data.Quoted 7 lines
Quoted 3 lines
Anyway, I haven't tried to create a small reproducer; I thought I would report it first and see if anything rang a bell. I tried this on a recent kernel and got the same issue.
I guess a combination of xmit rate, msg and buffer sizes and packet drops can lead to this delay. I've seen it happening, but didn't have the time to track it down back then.
There should be no packet drops. It's running across localhost, and flow control in the multiplex layer as it's set up for the tests limits the maximum outstanding data to 1024 bytes. With overhead and flow control messages it's maybe 1050 bytes of data that would ever be unacked. (It's not really testing throughput, it's testing the inner workings of the multiplexing protocol.)
If I understand this correctly per the RFCs, by default if there are two messages outstanding, it will send an sack immediately. Otherwise it waits 200ms. I'm guessing what is happening is that that SCTP sends a sack and then receives one more message and the upper layer stops because of flow control. Then the sack comes out in 200ms and things continue.
Quoted 8 lines
Quoted 3 lines
Actually, that still doesn't make sense. The lack of a sack shouldn't keep anything from sending unless the congestion window is closed, which shouldn't happen in this case. Am I correct?
I guess you still have the Nagle algorithm enabled. Try enabled the
SCTP_NODELAYsocket option: https://tools.ietf.org/html/rfc6458#section-8.1.5 at the sender side.It is enabled by default and will delay the sending of packet if they are not large enough (an implementation decision) and there is outstanding data.
Well, that was a surprise, disabling Nagle caused the problem to go away. Nagle generally doesn't make a difference when transferring lots of data.
I guess this is a bad interaction between Nagle and the SCTP sack algorithms. With TCP in my test, data is flowing both ways so data is always being acked, and Nagle is never significantly involved.
That's happening with SCTP, too, but in some situations a sack could be sent, you get one more packet sent, and that packet won't be acked until another packet is sent. So you have unacked data, and Nagle will hold any new data until it receives an ack for the outstanding packet. So you get stuck until the sack delay elapses. Bah.
This is sort of like the interaction between Nagle and TCP delayed ack. Which is sort of a bug, I guess, but well known. I have a number of ways to work around this issue, and I can document it so users can know.
Thanks for your help.
-corey
Quoted 28 lines
Best regards
MichaelQuoted 24 lines
-corey
Quoted 21 lines
So I think I can figure out how to make this work smoothly. I assume this doesn't happen in TCP because all packets carry an ack, and there is data flowing both ways all the time.
Thanks,
-corey
Quoted 12 lines
That said, remember that Linux SCTP doesn't support buffer auto-tuning. So considering you're running a stress test, you probably want to adjust them accordingly manually to avoid packet drops.
Marcelo
Quoted 5 lines
The library is at https://github.com/cminyard/gensio. I'd need to provide a patch for the tracing.
-corey
Quoted 84 lines · Corey Minyard
Quoted 81 lines · Michael Tuexen
Quoted 75 lines · Corey Minyard
Quoted 65 lines · Corey Minyard
Quoted 24 lines · Marcelo Leitner
Quoted 17 lines · Corey Minyard
Hi, it's me again with another strange issue. In case you didn't figure it out before, I'm working on a library that supports all different types of stream I/O, and SCTP is one supported building block. I noticed when I stacked a multiplexer layer on top of SCTP I started getting timeouts occasionally. It took a bit, but I finally realized that I was getting 200ms delays occasionally between sending a packet and receiving a packet. I verified this with a trace right at the
sctp_send()andsctp_recvmsg()calls. It doesn't seem to be regular in any way I can see, but it happens often enough to cause issues.If I replace the SCTP block with a TCP block, it works fine, and pretty much all the code is the same except where it does the read and write calls (including the
epoll()usage, and I have also switched toselect()and it has the same issue). The write calls don't seem to be the issue, I see two back-to-back writes a few microseconds apart and see a 200ms delay between the messages on the receive side.Quoted 5 lines
The test in question sets up two connections and does a big simultaneous bidirectional transfer. The test app has 4 threads waiting on
epoll()handling data and writing data.And the delay is always ~200ms. Which sounds suspicious.
Quoted 7 lines
That can be the delayed sack timer, in kernel.
/* Delayed sack timer - 200ms */
#defineSCTP_DEFAULT_TIMEOUT_SACK(200)You may tweak the sysctl net.sctp.sack_timeout and see if changes accordingly, or via
SCTP_PEER_ADDR_PARAMSor even enable immediate ack (by settingSPP_SACKDELAY_DISABLE)Ok, setting
SPP_SACKDELAY_DISABLEdoes make the problem go away.Quoted 15 lines
Quoted 8 lines
It's not using
sctp_sendv()at the moment, as the systems I'm running on don't have that yet. But the library does have support if it sees it is available.So I don't think it's my library; I've stared at it a bunch (and found a few other bugs) but I can't reconcile this one. There are no timers
Nice.
Quoted 4 lines
that would cause this in the code in question. Just basically an
epoll()call waiting on data and receive processing that is comparing data, along with write processing that is sending the same data.Quoted 7 lines
Quoted 3 lines
Anyway, I haven't tried to create a small reproducer; I thought I would report it first and see if anything rang a bell. I tried this on a recent kernel and got the same issue.
I guess a combination of xmit rate, msg and buffer sizes and packet drops can lead to this delay. I've seen it happening, but didn't have the time to track it down back then.
There should be no packet drops. It's running across localhost, and flow control in the multiplex layer as it's set up for the tests limits the maximum outstanding data to 1024 bytes. With overhead and flow control messages it's maybe 1050 bytes of data that would ever be unacked. (It's not really testing throughput, it's testing the inner workings of the multiplexing protocol.)
Quoted 6 lines
If I understand this correctly per the RFCs, by default if there are two messages outstanding, it will send an sack immediately. Otherwise it waits 200ms. I'm guessing what is happening is that that SCTP sends a sack and then receives one more message and the upper layer stops because of flow control. Then the sack comes out in 200ms and things continue.
Quoted 3 lines
Actually, that still doesn't make sense. The lack of a sack shouldn't keep anything from sending unless the congestion window is closed, which shouldn't happen in this case. Am I correct?
I guess you still have the Nagle algorithm enabled. Try enabled the
SCTP_NODELAYsocket option: https://tools.ietf.org/html/rfc6458#section-8.1.5 at the sender side.
Quoted 6 lines
Quoted 2 lines
It is enabled by default and will delay the sending of packet if they are not large enough (an implementation decision) and there is outstanding data.
Well, that was a surprise, disabling Nagle caused the problem to go away. Nagle generally doesn't make a difference when transferring lots of data.
Correct.
Quoted 4 lines
I guess this is a bad interaction between Nagle and the SCTP sack algorithms. With TCP in my test, data is flowing both ways so data is always being acked, and Nagle is never significantly involved.
TCP also uses delayed ACKs. However, bidirectional transfers are different from unidirectional ones.
Quoted 6 lines
That's happening with SCTP, too, but in some situations a sack could be sent, you get one more packet sent, and that packet won't be acked until another packet is sent. So you have unacked data, and Nagle will hold any new data until it receives an ack for the outstanding packet. So you get stuck until the sack delay elapses. Bah.
The same applies to TCP...
Quoted 6 lines
This is sort of like the interaction between Nagle and TCP delayed ack. Which is sort of a bug, I guess, but well known. I have a number of ways to work around this issue, and I can document it so users can know.
Thanks for your help.
You are welcome.
Best regards
Michael
Quoted 31 lines
-corey
Quoted 28 lines
Best regards
MichaelQuoted 24 lines
-corey
Quoted 21 lines
So I think I can figure out how to make this work smoothly. I assume this doesn't happen in TCP because all packets carry an ack, and there is data flowing both ways all the time.
Thanks,
-corey
Quoted 12 lines
That said, remember that Linux SCTP doesn't support buffer auto-tuning. So considering you're running a stress test, you probably want to adjust them accordingly manually to avoid packet drops.
Marcelo
Quoted 5 lines
The library is at https://github.com/cminyard/gensio. I'd need to provide a patch for the tracing.
-corey
From: Corey Minyard
Quoted 1 line
Sent: 14 July 2020 17:24
...
Quoted 14 lines
Well, that was a surprise, disabling Nagle caused the problem to go away. Nagle generally doesn't make a difference when transferring lots of data.
I guess this is a bad interaction between Nagle and the SCTP sack algorithms. With TCP in my test, data is flowing both ways so data is always being acked, and Nagle is never significantly involved.
That's happening with SCTP, too, but in some situations a sack could be sent, you get one more packet sent, and that packet won't be acked until another packet is sent. So you have unacked data, and Nagle will hold any new data until it receives an ack for the outstanding packet. So you get stuck until the sack delay elapses. Bah.
Quoted 3 lines
This is sort of like the interaction between Nagle and TCP delayed ack. Which is sort of a bug, I guess, but well known. I have a number of ways to work around this issue, and I can document it so users can know.
If you have an indefinite amount of data to send then even with Nagle enabled data should get sent when enough is queued to fill an ethernet frame.
The only proviso is that there is enough transmit window space. Doing that in SCTP is much more horrid than TCP.
I have seen issues caused by 0 RTT, slow start and delayed acks. But that was TCP with Nagle enabled.
If you want a 'full hands on debugging experience' enable ftrace on 'useful' function entry points (all syscall entry/exit events is a good start) and get the application to stop it on the 200ms timeout.
David
-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)