Re: [Resin-interest] Leak with 3.0.23 load balancing
477 of those threads had this stack: 0xff2c5870 _read + 0x8 0xe91db008 sock_read + 0x24 0xe91d8edc BIO_read + 0xcc 0xeaa49154 ssl3_read_n + 0x154 0xeaa492d0 ssl3_get_record + 0xe0 0xeaa49e78 ssl3_read_bytes + 0x1fc 0xeaa48dcc ssl3_read_internal + 0x44 0xeacb22a4 ssl_read + 0xa8 0xeb1e25d8 Java_com_caucho_vfs_JniStream_readNative + 0x54 Tom -Original Message- Subject: RE: Leak with 3.0.23 load balancing Thread count climbed steadily from a beginning of 70 to 300+ to finally 500+ when heap was exhausted. Tom ___ resin-interest mailing list resin-interest@caucho.com http://maillist.caucho.com/mailman/listinfo/resin-interest
Re: [Resin-interest] Leak with 3.0.23 load balancing
On Oct 3, 2008, at 7:03 AM, Tom Hintz wrote: 477 of those threads had this stack: 0xff2c5870 _read + 0x8 0xe91db008 sock_read + 0x24 0xe91d8edc BIO_read + 0xcc 0xeaa49154 ssl3_read_n + 0x154 0xeaa492d0 ssl3_get_record + 0xe0 0xeaa49e78 ssl3_read_bytes + 0x1fc 0xeaa48dcc ssl3_read_internal + 0x44 0xeacb22a4 ssl_read + 0xa8 0xeb1e25d8 Java_com_caucho_vfs_JniStream_readNative + 0x54 Hmm. Looks like the socket timeout isn't happening for some reason. I've filed a bug as http://bugs.caucho.com/view.php?id=2988 -- Scott Tom -Original Message- Subject: RE: Leak with 3.0.23 load balancing Thread count climbed steadily from a beginning of 70 to 300+ to finally 500+ when heap was exhausted. Tom ___ resin-interest mailing list resin-interest@caucho.com http://maillist.caucho.com/mailman/listinfo/resin-interest ___ resin-interest mailing list resin-interest@caucho.com http://maillist.caucho.com/mailman/listinfo/resin-interest
Re: [Resin-interest] Leak with 3.0.23 load balancing
On Oct 2, 2008, at 2:30 PM, Tom Hintz wrote: Thread count climbed steadily from a beginning of 70 to 300+ to finally 500+ when heap was exhausted. That's interesting. It would explain the memory error. If it's possible, a thread dump would really help. Also, the memory-free-max1M/memory-free-max setting would help. Its main purpose is to prevent the JVM from continually spinning when memory is almost exhausted. In that situation, a forced restart is better. One issue, though, is that 3.0 is frozen, so other than configuration tweaks, any Resin code changes would be in 3.1 or 3.2. -- Scott Tom ___ resin-interest mailing list resin-interest@caucho.com http://maillist.caucho.com/mailman/listinfo/resin-interest ___ resin-interest mailing list resin-interest@caucho.com http://maillist.caucho.com/mailman/listinfo/resin-interest