Tom Ridge

Under Linux, it is probably the case that the close call is guaranteed to close a valid fd, even if it subsequently reports an error such as EINTR. So we...

This is probably a FUSE question, rather than an ocamlfuse question, but perhaps someone can help... Suppose a user has an ocamlfuse filesystem mounted, and the user application calls write(somepath,buf,off,len)...

Victor Allombert reported on slack: --- I have two machines running the GC. The first, a powerful one, succeeded it's last GC (in a few minutes). However, the second one,...

irmin-pack: max memory usage 3x, main vs 3.3.1

8

This issue is based on #1959, but focusing solely on the increased memory usage. From the metrics listed in that issue: ``` | | 3.2-a | 3.2-b | 3.3.1-a |...

irmin-pack: reduced LRU cache hits, much increased disk reads (irmin main vs 3.3.1)

9

On 2022-07-07, Zach reported reduced LRU cache hits and much increased disk reads, when comparing irmin main branch and 3.3.1. Attached are the original stats.txt, and an annotated version highlighting...

irmin-pack: assert failure, pack_index.ml, line 108

6

From tezos-dev, devteam channel: https://tezos-dev.slack.com/archives/GB0UR34N8/p1658154805411149 ``` $ dune exec scripts/yes-wallet/yes_wallet.exe -- create from context ~/tezos/tezos-node-ghostnet/ in /tmp/yes-wallet --active-bakers-only --staking-share 75 Fatal error: exception File "src/irmin-pack/unix/pack_index.ml", line 108, characters 26-32: Assertion...

irmin-pack: possible increased disk usage for GC

4

It was reported that: _It seems that the disk footprint of the GC is around 40Gb for a recent context, while the context (~6 cycles) is ~37Gb (a total of...

Kind.of_magic: unexpected magic char '\000'

4

https://app.slack.com/client/T59LZHQ11/C596FGNUR/thread/C596FGNUR-1654186374.376829 From that link: Looks like we also have one v13.0 node with a strange error that I haven't seen before: ``` Jun 4 11:29:08.585 - node.validator.bootstrap_pipeline: in 79 steps...

irmin-pack: GC suffix copy in child process appears slow

5

In the child process, the copy to the new suffix should be iterated a number of times (or until there are few bytes left to copy). There is a branch...

layers/GC: make monitoring easier by having standard paths for the "current" sparse and suffix

2

(Suggested by Victor) At the moment we have directories like `sparse.1234` and `suffix.1234` where the suffix number corresponds to the generation (the generation gets incremented by one on each GC)....