<?xml version="1.0" encoding="utf-8"?>
<feed xml:lang="en" xmlns="http://www.w3.org/2005/Atom"><title>Recent changes to patches</title><link href="https://sourceforge.net/p/balance/patches/" rel="alternate"/><link href="https://sourceforge.net/p/balance/patches/feed.atom" rel="self"/><id>https://sourceforge.net/p/balance/patches/</id><updated>2017-12-10T00:29:45.080000Z</updated><subtitle>Recent changes to patches</subtitle><entry><title>Avoid deadlock</title><link href="https://sourceforge.net/p/balance/patches/7/" rel="alternate"/><published>2017-12-10T00:29:45.080000Z</published><updated>2017-12-10T00:29:45.080000Z</updated><author><name>Henrik Carlqvist</name><uri>https://sourceforge.net/u/henca/</uri></author><id>https://sourceforge.netbcf41249a326334e849033f1c9463695266287db</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;At high network load a deadlock migh happen leaving balance hanging trying to write data to the server while the server also hangs trying to write data back to balance. Killing the client  will leave the socket in a FIN_WAIT1 state. The problem is caused by the false assumption that only one file descriptor has data when select returns and this causes starvation of the server file descriptor. The attached patch fixes this by allways reading from all file descriptors that have new data after select returns. Maybe it would also be possible to fix this with better behaving servers and/or clients, but in my case this bug has been triggered with multiple levels of balance. That is balance has forwarded traffic to other load balancers using balance.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Connection count remains too high at fork failure</title><link href="https://sourceforge.net/p/balance/patches/6/" rel="alternate"/><published>2013-06-21T13:22:07.638000Z</published><updated>2013-06-21T13:22:07.638000Z</updated><author><name>Henrik Carlqvist</name><uri>https://sourceforge.net/u/henca/</uri></author><id>https://sourceforge.net8661bb76b0638e219a97346c84a385a7fd20ce6e</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;For each new connection the connection counter as well as the total connections counter for that channel is increased. Then a new process is forked and that process will restore the connection counter before exit when the channel is closed. However, if the fork fails, the counter showing the number of active connections will keep its higher value even though no connection remains as no connection was ever done. &lt;/p&gt;
&lt;p&gt;This patch restores the correct value for the active number of connections counter. It also restores the total number of connections counter. &lt;/p&gt;
&lt;p&gt;The patch was created from latest version 3.54 from www.inlab.de&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>MAXCHANNELS are per group, not a total number</title><link href="https://sourceforge.net/p/balance/patches/5/" rel="alternate"/><published>2011-05-07T17:38:39Z</published><updated>2011-05-07T17:38:39Z</updated><author><name>Henrik Carlqvist</name><uri>https://sourceforge.net/u/henca/</uri></author><id>https://sourceforge.net9e6b19926dc60259026e6b1629723c357fb643aa</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;The limitation of MAXCHANNELS (default 16) are per group (also max 16 groups) for a possible total of default max 16*16 = 256 channels evenly divided in 16 groups. However, in the code there is a check against no more than MAXCHANNELS total channels. This patch removes that faulty check.&lt;/p&gt;
&lt;p&gt;The code also contains the correct checks to avoid more than MAXCHANNELS channels in each group.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Fix typo in manpage balance(1)</title><link href="https://sourceforge.net/p/balance/patches/4/" rel="alternate"/><published>2007-02-23T11:52:29Z</published><updated>2007-02-23T11:52:29Z</updated><author><name>Anonymous</name><uri>https://sourceforge.net/u/userid-None/</uri></author><id>https://sourceforge.net9144317627368e3a27893e32452164f82fc20571</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;balance(1) of balance-3.35 misleadingly says, that "F" means "stay in foreground". However, The online help states that "f" is the right option.&lt;/p&gt;
&lt;p&gt;The attached patch fixes this typo by changing "F" to "f".&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>Enable TCP_NODELAY increased speed / less delay</title><link href="https://sourceforge.net/p/balance/patches/3/" rel="alternate"/><published>2005-08-30T13:16:04Z</published><updated>2005-08-30T13:16:04Z</updated><author><name>Wim Bonis</name><uri>https://sourceforge.net/u/bonis/</uri></author><id>https://sourceforge.net1f6690b570f0d6c7a6d47f2f862e9ca1f0f93213</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;We have found that the following patch speeds up the &lt;br /&gt;
the connection of a Terminal-Session (RDP).&lt;/p&gt;
&lt;p&gt;The Patch enables TCP_NODELAY on the incoming &lt;br /&gt;
and outgoing TCP channel.&lt;/p&gt;
&lt;p&gt;Can you please add this or a simmilar fix to the next &lt;br /&gt;
version?&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>timeouts: removed SIGALRM in favor of select()</title><link href="https://sourceforge.net/p/balance/patches/2/" rel="alternate"/><published>2003-05-07T22:39:34Z</published><updated>2003-05-07T22:39:34Z</updated><author><name>Dan Helfman</name><uri>https://sourceforge.net/u/witten/</uri></author><id>https://sourceforge.net896d07729ba1cf2f231fe2aa0c2e768d9c0de2d0</id><summary type="html">&lt;div class="markdown_content"&gt;&lt;p&gt;When using balance, I came across the following&lt;br /&gt;
problem: The connection failure timeout does not seem&lt;br /&gt;
to be honored when the host does not respond at all. I&lt;br /&gt;
tried listing a completely firewalled host as one of&lt;br /&gt;
the destination hosts on the balance command-line, and&lt;br /&gt;
balance will just block on a connection to it for much&lt;br /&gt;
longer than the given timeout parameter. So what I did&lt;br /&gt;
was remove the SIGALRM business on connect and instead&lt;br /&gt;
made the connect() call non-blocking and then added a&lt;br /&gt;
subsequent select() call so that the connect will&lt;br /&gt;
properly time out after the given number of seconds. In&lt;br /&gt;
some initial testing, it appears to work just great.&lt;/p&gt;&lt;/div&gt;</summary></entry><entry><title>RPM .spec file</title><link href="https://sourceforge.net/p/balance/patches/1/" rel="alternate"/><published>2000-11-01T00:44:05Z</published><updated>2000-11-01T00:44:05Z</updated><author><name>Sean Reifschneider</name><uri>https://sourceforge.net/u/jafo/</uri></author><id>https://sourceforge.net09f147b1c883a184296c26cfc0d37d482740c5b5</id><summary type="html"/></entry></feed>