curvefs/client: fix the data iteration error when rpc retry. #1474
Conversation
| std::vector<std::list<uint64_t>> *inodeGroups) { | ||
| std::unordered_map<uint32_t, std::list<uint64_t>> groups; |
There was a problem hiding this comment.
for most cases, std::vector is always preferred, std::list/map/set are all node-based containers, they're not cache-friendly.
for std::vector, another tip is if you know the number of elements that it will store, you can call vec.reserve firstly to allocate enough memory.
| if (ret) { | ||
| for (const auto &it : groups) { |
There was a problem hiding this comment.
if (!ret) {
return false;
}
for (...| for (const auto &id : it) { | ||
| request.add_inodeid(id); | ||
| } |
There was a problem hiding this comment.
might be a shortcut, *request.mutable_inode() = { it.begin(), it.end() }
| for (const auto &id : it) { | ||
| request.add_inodeid(id); | ||
| } | ||
| if (request.inodeid_size() <= 0) { |
There was a problem hiding this comment.
should you check whether it is empty or not at the beginning? otherwise L537 will be core dumped.
| << ", errmsg = " << MetaStatusCode_Name(ret); | ||
| } else if (response.attr_size() > 0 && | ||
| response.has_appliedindex()) { | ||
| auto retAttr = response.attr(); |
There was a problem hiding this comment.
avoid unnecessary copy, use reference here.
| auto retXattr = response.xattr(); | ||
| for_each(retXattr.begin(), retXattr.end(), | ||
| [&](XAttr &a) { xattr->push_back(a); }); |
There was a problem hiding this comment.
auto* attrs = response.mutable_xattr();
xattr->insert(xattr->end(),
std::make_move_iterator(attrs->begin()),
std::make_move_iterator(attrs->end()));might be a better way to just move xattrs in response to the destination.
| uint64_t inodeId = *it.begin(); | ||
| auto task = RPCTask { | ||
| metaserverClientMetric_->batchGetXattr.qps.count << 1; | ||
| LatencyUpdater updater( | ||
| &metaserverClientMetric_->batchGetXattr.latency); |
There was a problem hiding this comment.
the overall process is still serial, can we send requests parallelly?
There was a problem hiding this comment.
Yes, add a TODO and improve this in later pr.
What problem does this PR solve?
Issue Number: #xxx
#1427
Problem Summary:
The before logic will crash when rcp retry, the iter will forward when retry but this is not expected.
What is changed and how it works?
What's Changed:
How it Works:
Side effects(Breaking backward compatibility? Performance regression?):
Check List