Hi, sorry, I was a bit busy. So I did put back the tcp_outgoing_address. Will I have to use it forever or just for testing? Does it have to be an IP different than the regular IP? Again, post a bit long... squid1: 192.168.17.11 / tcp_ougoing 192.168.17.15 squid2: 192.168.17.12 / tcp_ougoing 192.168.17.16 squid3: 192.168.17.13 / tcp_ougoing 192.168.17.17 squid4: 192.168.17.14 / tcp_ougoing 192.168.17.18 digest_rebuild_period 10 minutes I start the squids after an rm of the cache_dirs. ### I ask squid1 for object1 squid1: 1216303244.157 2 192.168.17.11 TCP_MISS/200 2901 GET http://192.168.16.23/index.html - FIRST_UP_PARENT/apache text/html squid2: 1216303243.855 0 192.168.17.11 UDP_MISS/000 52 ICP_QUERY http://192.168.16.23/index.html - NONE/- - 1216303244.155 0 192.168.17.15 TCP_MEM_HIT/200 477 GET internal://squid2/squid-internal-periodic/store_digest - NONE/- application/cache-digest squid3: 1216303244.053 0 192.168.17.11 UDP_MISS/000 52 ICP_QUERY http://192.168.16.23/index.html - NONE/- - squid4: 1216303243.314 0 192.168.17.11 UDP_MISS/000 52 ICP_QUERY http://192.168.16.23/index.html - NONE/- - Then, 1mn later, squid3: 1216303304.155 0 192.168.17.15 TCP_MEM_HIT/200 477 GET internal://squid3/squid-internal-periodic/store_digest - NONE/- application/cache-digest Then, 1mn later, squid4: 1216303364.156 0 192.168.17.15 TCP_MEM_HIT/200 477 GET internal://squid4/squid-internal-periodic/store_digest - NONE/- application/cache-digest ### I ask squid3 for object2 squid1: 1216303708.034 0 192.168.17.13 UDP_MISS/000 55 ICP_QUERY http://192.168.16.23/img/spain.gif - NONE/- - 1216303708.324 0 192.168.17.17 TCP_MEM_HIT/200 477 GET internal://squid1/squid-internal-periodic/store_digest - NONE/- application/cache-digest squid2: 1216303707.821 0 192.168.17.13 UDP_MISS/000 55 ICP_QUERY http://192.168.16.23/img/spain.gif - NONE/- - squid3: 1216303708.330 6 192.168.17.13 TCP_MISS/200 7300 GET http://192.168.16.23/img/spain.gif - FIRST_UP_PARENT/apache image/gif squid4: 1216303708.242 0 192.168.17.13 UDP_MISS/000 55 ICP_QUERY http://192.168.16.23/img/spain.gif - NONE/- - Then, 1mn later, squid2: 1216303768.325 0 192.168.17.17 TCP_MEM_HIT/200 477 GET internal://squid2/squid-internal-periodic/store_digest - NONE/- application/cache-digest Then, 1mn later, squid4: 1216303828.326 0 192.168.17.17 TCP_MEM_HIT/200 477 GET internal://squid4/squid-internal-periodic/store_digest - NONE/- application/cache-digest ### Etc... Now all objects are randomly spread each on 1 unique squid. ### And in the meantime: squid2: 1216303945.412 0 192.168.17.18 TCP_MEM_HIT/200 477 GET internal://squid2/squid-internal-periodic/store_digest - NONE/- application/cache-digest squid3: 1216303950.276 0 192.168.17.16 TCP_MEM_HIT/200 477 GET internal://squid3/squid-internal-periodic/store_digest - NONE/- application/cache-digest Then, squid3: 1216304005.426 0 192.168.17.18 TCP_MEM_HIT/200 477 GET internal://squid3/squid-internal-periodic/store_digest - NONE/- application/cache-digest squid4: 1216304010.277 0 192.168.17.16 TCP_MEM_HIT/200 477 GET internal://squid4/squid-internal-periodic/store_digest - NONE/- application/cache-digest ### I do another round... ### Works well (3 UDP query and 1 HIT, unless the squid has it already) up to object5... ### Ask squid2 for object6 (squid1 has it, no UDP traffic in logs): squid2: 1216304170.370 1 192.168.17.12 TCP_MISS/200 9042 GET http://192.168.16.23/img/sweden.gif - FIRST_UP_PARENT/apache image/gif squid4: 1216304170.369 0 192.168.17.16 TCP_MISS/504 1585 GET http://192.168.16.23/img/sweden.gif - NONE/- text/html Then works again for a few objects... Then not: ask squid2 for object (no UDP traffic in logs). squid2: 1216304475.185 0 192.168.17.12 TCP_MISS/200 8486 GET http://192.168.16.23/img/polska.gif - CD_SIBLING_HIT/squid3 image/gif squid3: 1216304475.184 147 192.168.17.16 TCP_MEM_HIT/200 8392 GET http://192.168.16.23/img/polska.gif - NONE/- image/gif ### I do another round... ### ask squid2 for object1 (no UDP traffic)... squid1: 1216304762.804 0 192.168.17.16 TCP_MEM_HIT/200 3018 GET http://192.168.16.23/index.html - NONE/- text/html squid2: 1216304762.804 0 192.168.17.12 TCP_MISS/200 3112 GET http://192.168.16.23/index.html - CD_SIBLING_HIT/squid1 text/html ### ask squid4 for object2 (no UDP traffic)... squid2: 1216304825.551 0 192.168.17.18 TCP_MISS/504 1583 GET http://192.168.16.23/img/spain.gif - NONE/- text/html squid4: 1216304825.552 1 192.168.17.14 TCP_MISS/200 7300 GET http://192.168.16.23/img/spain.gif - FIRST_UP_PARENT/apache image/gif In fact, squid3 had it... I did no purges at all. And above we can see 192.168.17.18 TCP_MEM_HIT/200 477 GET internal://squid3/squid-internal-periodic/store_digest - NONE/- application/cache-digest The digest exchange was after squid3 cached spain. The 2 following objects were also missed and fetched from apache again. Then 3 UDP requests ok (so I guess no digest used). Then 1 failed. Then 2 direct hits. Then 1 good digest hit. Are these digests "collisions"? Did several more rounds, still with no purges, and several objects were again fetched from apaches. As usual, with no-digest, it works fine. Thx, JD
