Attachment 242124 Details for Bug 271367 – git(1) diff against the ports tree

[patch] git(1) diff against the ports tree

0001-sysutils-py-salt-workaround-a-hanging-minion-issue.patch (text/plain), 3.35 KB, created by Kyle Evans on 2023-05-11 18:11:21 UTC

(hide)

Description:

Filename:

MIME Type:

Creator: Kyle Evans

Created: 2023-05-11 18:11:21 UTC

Size: 3.35 KB

patch

obsolete

>From 6d8f2315c8f0d03897ae48cb52e1324e13e8081a Mon Sep 17 00:00:00 2001
>From: Kyle Evans <kevans@FreeBSD.org>
>Date: Thu, 11 May 2023 13:03:00 -0500
>Subject: [PATCH] sysutils/py-salt: workaround a hanging minion issue
>MIME-Version: 1.0
>Content-Type: text/plain; charset=UTF-8
>Content-Transfer-Encoding: 8bit
>
>From the patch description:
>
>--
>Workaround for https://github.com/saltstack/salt/issues/64074
>
>As of salt v3005, it's possible for a minion process to get
>irrecoverably wedged due to some restructuring in IPC.  Namely, payloads
>that are to be reported back to the master are written over a UNIX
>socket by the very same thread that is expected to drain the socket's
>receive buffer.
>
>Smaller payloads are fine, but anything that exceeds FreeBSD's default
>receive buffer size (8KB) results in the process hanging in socket write
>for some buffer space that will never come.  The most obvious way to
>identify a hung minion is a lingering zombie child, but it will more
>generally stop responding to salt-master.
>--
>
>The workaround is to simply bump the receive buffer size so that the
>vast majority of payloads won't trigger it.  This code likely requires
>some restructuring upstream to avoid writing to a pipe that the same
>thread is responsible for draining.
>
>Submitted by:	Johan SÃ¶llvander <jsollvander_axcient.com>
>---
> sysutils/py-salt/Makefile                     |  2 +-
> .../files/patch-salt_ext_tornado_iostream.py  | 23 +++++++++++++++++++
> 2 files changed, 24 insertions(+), 1 deletion(-)
> create mode 100644 sysutils/py-salt/files/patch-salt_ext_tornado_iostream.py
>
>diff --git a/sysutils/py-salt/Makefile b/sysutils/py-salt/Makefile
>index b32abac6a4de..25bd78b9f2e3 100644
>--- a/sysutils/py-salt/Makefile
>+++ b/sysutils/py-salt/Makefile
>@@ -1,6 +1,6 @@
> PORTNAME=	salt
> PORTVERSION=	3005.1
>-PORTREVISION=	1
>+PORTREVISION=	2
> CATEGORIES=	sysutils python
> MASTER_SITES=	PYPI
> PKGNAMEPREFIX=	${PYTHON_PKGNAMEPREFIX}
>diff --git a/sysutils/py-salt/files/patch-salt_ext_tornado_iostream.py b/sysutils/py-salt/files/patch-salt_ext_tornado_iostream.py
>new file mode 100644
>index 000000000000..2ea532b1e761
>--- /dev/null
>+++ b/sysutils/py-salt/files/patch-salt_ext_tornado_iostream.py
>@@ -0,0 +1,23 @@
>+Workaround for https://github.com/saltstack/salt/issues/64074
>+
>+As of salt v3005, it's possible for a minion process to get irrecoverably
>+wedged due to some restructuring in IPC.  Namely, payloads that are to be
>+reported back to the master are written over a UNIX socket by the very same
>+thread that is expected to drain the socket's receive buffer.
>+
>+Smaller payloads are fine, but anything that exceeds FreeBSD's default receive
>+buffer size (8KB) results in the process hanging in socket write for some buffer
>+space that will never come.  The most obvious way to identify a hung minion is
>+a lingering zombie child, but it will more generally stop responding to
>+salt-master.
>+
>+--- salt/ext/tornado/iostream.py.orig	2022-11-02 13:25:04 UTC
>++++ salt/ext/tornado/iostream.py
>+@@ -1116,6 +1116,7 @@ class IOStream(BaseIOStream):
>+             future = self._connect_future = TracebackFuture()
>+         try:
>+             self.socket.connect(address)
>++            self.socket.setsockopt(socket.SOL_SOCKET, socket.SO_SNDBUF, 524288)
>+         except socket.error as e:
>+             # In non-blocking mode we expect connect() to raise an
>+             # exception with EINPROGRESS or EWOULDBLOCK.
>-- 
>2.40.0
>

Actions: View | Diff

Attachments on bug 271367: 242124